West African Journal of Industrial & Academic Research ISSN: 2276-9129

Transcription

ISSN: 2276-9129
West African Journal of Industrial & academic research
August 31, 2012 Vol. 4 No.1
West African Journal of Industrial
& Academic Research
www.wajiaredu. com email: info@ wajiaredu.com
Evaluation And Comparison Of The Principal Component Analysis (PCA) and Isometric Feature Mapping
(Isomap) Techniques On Gas Turbine Engine Data
Uduak A.Umoh, Imoh J.Eyoh and Jeremiah E. Eyoh
3
On the Probability Density Functions of Forster-Greer-Thorbecke (FGT) Poverty Indices
Osowole, O. I., Bamiduro, T.A
10
Comparison of Three Criteria for Discriminant Analysis Procedure
Nwosu, Felix D., Onuoha, Desmond O. and Eke Charles N.
17
A Computational Analysis of the Negative Impact of Cigarette Smoking on Human Population In Imo State
Ekwonwune E, Osuagwu O.E, Edebatu D
30
An Application of Path Sharing To Routing For Mobile Sinks In Wireless Sensor Networks
Okafor Friday Onyema, Fagbohunmi Griffin Siji
42
Expert System for Diagnosis of Hepatitis B
Ibrahim Mailafiya, Fatima Isiaka
57
A Comparative Performance Analysis of Popular Internet Browsers in Current Web Applications
Boukari Souley, Amina S. Sambo
69
Adjusting for the Incidence of Measurement Errors in Multilevel Models Using Bootstrapping and Gibbs Sampling Techniques
Imande, M.T and Bamiduro, T.A
79
Design and Implementation of an M/M/1 Queuing Model Algorithm and its Applicability in Remote Medical Monitoring
.
Ifeoma Oji and Osuagwu O.E
IISTRD
94
Classification of Implemented Foreign Assisted Projects into Sustainable And Non-sustainable Groups: A Discriminant Analysis Approach:
Iwuagwu Carmelius Chibuzo
110
A Study on the Evaluation of Industrial Solid Waste Management Approaches in Some Industries in Aba, South Eastern Nigeria
Ajero, C.M.U and Chigbo,U.N
114
Deploying Electronic Roadside Vehicle Identification Technology to Intercept Small Arms and Ammunition on Nigeria Roads
Akaenyi, I.W, Osuagwu O.E
126
Statistical Analysis of Deviance among Children in Makurdi Metropolis
Kembe, M.M and Kembe, E.M
143
A Profile Analysis on the Effectiveness of Two kinds of Feeds on Poultry Birds.
Onuoha, Desmond O. and Opara Pius N
155
Information and Communication Technology (Ict) Integration Into Science, Technology, Engineering And Mathematic (Stem) In Nigeria
A.A. Ojugo., A. Osika., I.J.B. Iyawa and R.O. Yerokun (Mrs.)
169
Comparative Analysis of the Functions 2n, n! and nn
Ogheneovo, E. E.; Ejiofor, C. and Asagba, P. O
179
Implementation of A Collaborative E-Learning Environment On A Linux Thin-Client System
Onyejegbu L. N. and Ugwu C.
185
An assessment of Internet Abuse in Nigeria
M.E Ezema, H.C. Inyama
191
Editor-in-Chief: Prof. O. E. Osuagwu, FNCS, FBCS
Vol.
Vol.4
ol.4 No.1. August 2012
& Academic Research
Publications Office:
9-14 mbonu Ojike Street
Ikenegbu, Owerri, Nigeria
International office::
255 North D Street
San Bernardino, CA 92401
Tel: 234 81219 41139
www.wajiaredu.com
Editor-in-Chief:
Editorial Board:
909.884.9000
Prof. Oliver E. Osuagwu, PhD, FNCS, FBCS CITP, MIEEE, MACM
Prof Tony B.E. Ogiemien, PhD, BL, (USA), Engr. Prof E. Anyanwu, Ph.D, FNSE, Prof. G. Nworuh, PhD, Dr. B. C. Ashiegbu, PhD, Prof.
C.O.E. Onwuliri, PhD, FAS , Prof .E. Emenyionu, PhD, (Connecticut USA,) Prof. E.P. Akpan, Ph.D, Engr. Prof. C.D. Okereke, Ph.D, Prof. B.E.B. Nwoko, Ph.D, Prof. N.N.
Onu, PhD, Prof M.O. Iwuala, PhD, Prof C.E.Akujo, PhD, Prof. G. Okoroafor, PhD, Prof Leah Ojinna, Ph.D (USA), Prof. O. Ibidapo-Obe, PhD, FAS., Prof. E. Adagunodo,
PhD, Prof. J.C .Ododo, PhD, Dan C. Amadi, PhD(English), Prof.(Mrs) S.C. Chiemeke, PhD, Prof (Mrs) G. Chukwudebe,PhD, FNSE, Dr. E.N.C. Okafor, PhD, Dr (Mrs) I.
Achumba, Dr. T. Obiringa, PhD, Dr. S. Inyama, PhD, Prof. C. Akiyoku, PhD, Prof. John Ododo, PhD, Prof. E. Nwachukwu, Ph.D, FNCS, Dr. S. Anigbogu, PhD,FNCS, Prof. H.
Inyama, PhD, FNSE .Prof. B.N.. Onwuagba, PhD, Prof J.N. Ogbulie, PhD
Published by:
Olliverson Industrial Publishing House
The Research & Publications Division of Hi-Technology Concepts (WA) Ltd
For The
International Institute for Science, Technology Research & Development, Owerri,
Nigeria & USA
All rights of publication and translation reserved. Permission for the reproduction of text and illustration
should be directed to the Editor-in-Chief @ OIPH, 9-14 Mbonu Ojike Street, Ikenegbu, Owerri, Nigeria or
via our email address or the international office for those outside Nigeria
© International Institute for Science, Technology Research &
Development, Owerri, Nigeria
West African Journal of Industrial and Academic Research Vol.4 No. 1 August 2012
2
Evaluation and Comparison of the Principal Component Analysis (PCA) and Isometric
Feature Mapping (Isomap) Techniques on Gas Turbine Engine Data
Uduak A.Umoh+, Imoh J.Eyoh+ and Jeremiah E. Eyoh*
+Department of Computer Science,
University of Uyo, Uyo, Akwa Ibom State, Nigeria
*Department of Turbo Machinery (Reliability and Maintenance),
Exxon Mobile, QIT, Eket, Akwa Ibom State, Nigeria
Abstract
This paper performs a comparative analysis of the results of PCA and ISOMAP for the purpose of reducing or
eliminating erratic failure of the Gas Turbine Engine (GTE) system. We employ Nearest-neighbour
classification for GTE fault diagnosis and M-fold cross validation to test the performance of our models.
Comparison evaluation of performance indicates that, with PCA, 80% of good GTE is classified as good GTE,
77% of the average GTE is classified as average GTE and 67.6% of bad GTE is classified as bad GTE. With
ISOMAP, 67% of good GTE is classified as good GTE, 70.8% of the average GTE is classified as average GTE
and 81% of bad GTE is classified as bad GTE. PCA produces 26% error rate with nearest neighbour
classification and 17% error rate with M-fold cross validation. While ISOMAP produces 35% error rate with
nearest neighbour classification, and 26.5% error rate with M-fold cross validation. Results indicate that PCA
is more effective in analyzing the GTE data set, giving the best classification for fault diagnosis. This enhances
the reliability of the turbine engine during wear out phase, through predictive maintenance strategies.
_______________________________________________________________________________
1.0 Introduction
Maintenance of complex engineering systems
such as GTE has posed a serious challenge to
systems engineers, as this affects the GTE
subsystems and entire system reliability and
performance. Monitoring the health of a system is
part of the predictive maintenance approach that
seeks to extend the reliability and life of the system.
Principal Component Analysis (PCA) and Isomeric
Feature Mapping (ISOMAP) are dimensionality
reduction techniques employed to transform a highdimensional data space to a low-dimensional space
with information and local structure of the data set
being preserved as much as possible. Principal
Components Analysis, PCA has been proven to be
good in transforming high dimensional linear data
set to lower dimensional space, with much lose of
information contained in the original data.
Applying linear techniques of dimensionality
reduction to a nonlinear data such as GTE data set
is sure not going to give a much success story as
when linear techniques are applied to a linear data
set. Isometric Feature Mapping, ISOMAP is a
nonlinear dimensionality reduction method that
maps from the high dimensional space to a lowdimensional Euclidean feature space. Also, the
projected observation with reduced dimensions
preserves as much as possible the intrinsic metric
structure of the observation [9]. In this work, we
evaluate
and
compare
analyzed
signal
characteristics and extracted features based on PCA
and ISOMAP data-based analysis techniques. We
explore Matlab and C++ programming tools for the
implementation. .
2.0
Literature Review
Gas turbine engines have proven to be very
efficient and are widely used in many industrial and
engineering systems. They are used in systems such
as Aircrafts, Electrical power generation Systems,
Trains, Marine vessels, as drivers to industrial
equipment such as high capacity compressors and
pumps. In most cases, areas of application of gas
turbine engines are safety critical which require
very high reliability and availability of these
systems. To maintain high system reliability and
availability, critical system parameter variables
3
such as engine vibration, bearing temperature, lube
oil pressure, etc, must be continuously monitored
for prompt detection of deviation from normal
operation values. To design a system for high
reliability means, increasing the cost of the system
and its complexity [4]. More so, monitoring, control
and protection subsystems of the Gas Turbine
Engines further add more cost and complexity to
the overall system. The application of a classical
maintenance approaches has been proven over the
years, to be unsuitable for engineering systems such
as Gas turbine engines [7] [6]. The health state of a
GTE is determined by its functional state or
characteristics of the parameter variables.
Depending on the characteristics of these parameter
variables, the GTE health state can be in a
particular state [7]. In PCA, data can be
transformed to a new set of coordinates or variables
that are a linear combination of the original
variables [8]. Researchers have developed various
systems’ health condition monitoring strategies in
which the state of the system is expected to operate
under designed operating conditions. Thus,
condition based predictive maintenance has
significant cost reduction implications [7].
The health state of a GTE is determined by its
functional state or characteristics of the parameter
variables. Depending on the characteristics of these
parameter variables, the GTE health state can be in
any of the following states [7]. Basic fault models
are due [6] [7] [1] [10]. Most of the turbine engine
diagnostic and prognostic systems are based on
model-based and, or knowledge-based approaches,
in which artificial neural networks techniques are
used. Some of the disadvantages of this approach
are that it adds more cost to the system life cycle
and further physical and architectural structure of
this complex system greatly reduces the reliability
of the entire system [5].
3.
Research Methodology
Data-based health condition monitoring of GTE
employs dimensionality reduction techniques to
analyze the systems parameter variable data in
order to extract hidden features which are useful in
fault detection and diagnosis. This is achieved by
exploring different data classification techniques
for fault diagnosis. We first applied PCA to the
EngData training set to project the high
dimensional nonlinear data to a low-dimensional
subspace [2]. The low dimensional data obtained
shows that over 90 % of the information contained
in the original high dimensional data is found in
just the first ten principal component of the
analysis. The ISOMAP technique, which is
nonlinear method, is also applied to the data and the
reduced dimensional data is further analyzed [3].
We evaluate, and compare the results of PCA and
ISOMAP on the training data, using nearestneighbour classification and cross validation
techniques.
4.0
Performance Evaluation of PCA and
ISOMAP
a.
PCA
Though many techniques are available to test the
performance of the data model developed using
PCA, its performance is in a way, dependent on the
nature of the data set being analyzed. PCA will
perform much better analysis if the data set is
normally distributed around the median. Before the
PCA is applied on the data, it is first of all preprocessed to standardize the data for better results.
The data was standardized to have zero mean, unit
standard deviation and unity variance [2].
The analysis of the GTE training data set
produces 15 PCs, Eigen values as shown on Table
1. The low-dimensional basis based on the principal
components minimizes the reconstruction error,
which is given by:
℮ = ║x - x̂ ║
(1)
This error e can be rewritten as;
℮=
(2)
Where N = 98; K = 10, 11, 12, 13, 14, 15.
Throughout the analysis of this work, K is chosen
to be 12.
Calculating error when k = 10 is as follows;
℮ = ½ (98 – 88.8294)
℮ = 4.5853
For K = 12;
℮ = ½ (98 –90.8989)
℮ = 3.551
The residual error is relatively small as can be
seen from the calculation when K = 12, as used in
this analysis. This also indicates that PCA has been
able to analyze the data comparatively well, though
the GTE data is nonlinear and the distribution of the
data is not perfectly around the median.
4
The classification of GTE classes is shown in
Table 2. Here 80% of good GTE are classified as
good GTE, 77% of the average GTE was classified
as average GTE and 67.6% of bad GTE are
classified as bad GTE. No bad GTE are classified
as good GTE and no good GTE are classified as
bad GTE. This achievement by PCA is vey
commendable as it is very paramount in safety
critical systems such as GTE.
We employ cross validation method to test the
performance of the data-based model developed
using PCA.
Table 1 showing 15 PCs, Eigen values
Principal
Components (PCs)
PC#1
PC#2
PC#3
PC#4
PC#5
PC#6
PC#7
PC#8
PC#9
PC#10
PC#11
PC#12
PC#13
PC#14
PC#15
Rival (latent)
36.6533
13.9509
8.5086
7.2647
6.4723
4.8586
3.7902
3.2723
2.3949
1.6638
1.2393
0.9210
0.8787
0.7817
0.7240
Camus
Rival
36.6533
50.6042
59.1128
66.3774
72.8498
77.7083
81.4985
84.7708
87.1657
88.8294
90.0688
90.9898
91.8685
92.6502
93.3743
of
Colum of Rival (%)
37.4013
51.6369
60.3191
67.7321
74.3365
79.2942
83.1617
86.5008
88.9445
90.6423
91.9069
92.8467
93.7434
94.5410
95.2799
Table 2 Percentage of classification result with PCA
KNOWN
CLASSIFICATION
PREDICTED CLASSIFICATION
Good GTE
(class 1)
Good
GTE
12 (80%)
(class 1)
Average
GTE
11
(class 2)
Bad
GTE
0
(class 3)
Average
(class 2)
GTE
Bad
GTE
(class 3)
3
0
37 (77%)
0
12
25 (67.6%)
Total number of test cases = 100
Total number of Good GTE = 15; percentage of good GTE classification = 80%
Total number of Average GTE = 48; percentage of average GTE classification = 77%
Total number of Bad GTE = 37; percentage of bad GTE classification = 67.6%
Table 2 shows that 12 good GTE out of 15 were
classified as good GTE, 3 good GTE out of 15 were
classified as average GTE and no good GTE was
classified as bad GTE. Also, from the table, it can
be seen that no bad GTE was classified as good
GTE. This is very reasonable for safety critical
system such as GTE.
Despite the fact that the GTE data set is noisy
and nonlinear, the result from PCA is very
impressive because of the following achievements:
The residual error is reasonably small. The high
dimensional data space is projected to low-
dimensional subspace without much lost of
information contained in the original data. 80% of
good GTE was classified as good GTE, 77% of the
average GTE is classified as average GTE and
67.6% of bad GTE is classified as bad GTE. No bad
GTE is classified as good GTE and no good GTE
was classified as bad GTE. This achievement by
PCA is vey commendable as it is very paramount in
safety critical systems such as GTE. The cross
validation of the training model of the data base
also recorded an impressive result; that is 83% of
the training data model is classified while only 17%
5
The performance of ISOMAP can be evaluated
using nearest neighbour classification of the test
data set and cross validation of the training data set.
In this project work, the performance of ISOMAP
is seriously affected by the choice of
neighbourhood factor, k for the algorithm. This
may be due to the nature of the data set. The
neighbourhood factor above 8 gives a
comparatively bad result while a value of k below 7
leads to discontinuity and the Y.index (which
contains the indices of the points embedded),
produced is less than 98 indices. When k = 6 or 5
was used, the Y.index was 35 and k = 3 gave much
lower indices. This made the ISOMAP analysis
limited to only neighbourhood factor values. That is
7 or 8. Table 3 presents percentage of classification
result with ISOMAP when k = 7. Table 4 shows
percentage of classification result with ISOMAP
when K = 8
of the training data model is misclassified.
Therefore PCA has been able to detect 80% of the
good GTE, 77% of the average GTE and 67.6% of
the bad GTE, though PCA is not always an optimal
dimensionality
reduction
procedure
for
classification purposes.
b.
ISOMAP
As stated in the case of PCA, the effectiveness or
performance of ISOMAP depends on the nature of
the data set. ISOMAP give a better result for
manifolds of moderate dimensionality, since the
estimates of manifold distance for a given graph
size degrades as the dimensionality increases. The
data set whose classes or features are sparsely
distributed without defined uniformity, such as
engineering data obtained from practical systems,
may not give a better result when analyzed using
ISOMAP.
Table 3 Percentage of classification result with ISOMAP when K = 7
KNOWN
CLASSIFICATION
Good
(class 1)
Average
(class 2)
Bad
(class 3)
Good GTE Average GTE
(class 1)
(class 2)
GTE 0 (0%)
14
Bad
GTE
(class 3)
1
GTE
0
33(68.75%)
15
GTE
0
5
32 (86%)
Total number of Good GTE = 15; percentage of good GTE classification = 0%
Total number of Average GTE = 48; percentage of average GTE classification = 68.75%
Total number of Bad GTE = 37; percentage of bad GTE classification = 86%
Table 4 Percentage of classification result with ISOMAP when K = 8
KNOWN
CLASSIFICATION
Good GTE
(class 1)
Good
GTE
1 (6.7%)
(class 1)
Average
GTE
3
(class 2)
Bad
GTE
0
(class 3)
Average
(class 2)
GTE
Bad
GTE
(class 3)
14
0
34 (70.8%)
11
7
30 (81%)
Total number of Good GTE = 15; percentage of good GTE classification = 6.7%
Total number of Average GTE = 48; percentage of average GTE classification = 70.8%
Total number of Bad GTE = 37; percentage of bad GTE classification = 81%
6
With K = 8, ( though, even number is not a good
choice for K), the classification gives a slightly
good result as no good GTE was classified as bad
GTE and no bad GTE is classified as good GTE. It
is still not generally good approach because 14 out
of 15 good GTE are classified as average GTE.
Figure 1 presents Residual Variance vs Isomap
dimensionality with K = 7. ISOMAP technique
applied to GTE data set is able to correctly
recognize its intrinsic three-dimensionality as
indicated by the arrow in Figure 1.
0.35
plot of variance vs dimensions
when k = 7
0.3
Residual variance
0.25
0.2
The knee point
0.15
0.1
0.05
0
0
2
4
6
8
Isomap dimensionality
10
12
Fig. 1: Residual Variance vs Isomap dimensionality with K = 7
Other achievement by ISOMAP of the GTE data
set includes the following:
ISOMAP generated a two-dimensional embedding
with a neighbourhood graph which gives a visual
information or characteristic of the data set. This is
helpful in studying the geometrical structure of the
GTE data. Also, the ISOMAP analysis preserves
information contained in the data and the local
structure of the data.
With k = 8, ISOMAP is achieve 6.7% of good
GTE is classified as good GTE, 70.8% of the
average GTE is classified as average GTE and 81%
of bad GTE is classified as bad GTE. No bad GTE
is classified as good GTE and no good GTE is
classified as bad GTE. This achievement is
reasonably good as no it is important in safety
critical systems such as GTE. But the system
availability and productivity is affected as over
93% of good GTE is classified as average GTE.
The cross validation of the training model of the
data base using ISOMAP also recorded an
impressive result; that is 73.5% of the training data
model was classified while only 26.5% of the
training data model was misclassified.
5. Comparison of PCA and ISOMAP Analysis
Results.
PCA and ISOMAP are dimensionality reduction
techniques employed to transform a highdimensional data space to a low-dimensional space
with information and local structure of the data set
being preserved as much as possible. Both
techniques use the number of significant Eigen
values to estimate the dimensionality.
ISOMAP is a graph-based, spectral, nonlinear
method of dimensionality reduction approach with
no local optima. It is parametric, non-iterative,
polynomial time procedure which guarantees global
optimality. PCA is non-parametric, linear method in
which the direction of the greatest variance is the
eigenvector corresponding to the largest Eigen
values of the data set. PCA is guarantee to recover
the correct or true structure of the linear manifolds
while ISOMAP is guaranteed to recover the correct
or true dimensionality and geometrical structure of
a large class of non linear manifolds as shown in
Figures 4 and 5. The knee point in the Figure 4
indicates the true dimensionality of the manifold,
while in Figure 5; the PCA cannot recover the
correct dimensionality. In this work, when the two
methods are applied on the GTE data set, the results
show that PCA best analyzed the data than
ISOMAP. Table 5 compares the results obtained
from both methods. Thus PCA performance for this
analysis is better than ISOMAP. Figure 5 shows
comparison evaluation of PCA and ISOMAP
performance of the training data using nearestneighbour classification and cross validation. PCA
produced 26% error rate with nearest neighbour
classification, and 17% error rate with M-fold cross
validation. ISOMAP produced 35% error rate with
nearest neighbour classification, and 26.5% error
rate with M-fold cross validation.
0.35
plot of variance vs dimensions
when k = 7
0.3
Residual variance
0.25
0.2
The knee point
0.15
0.1
0.05
0
0
2
4
6
8
Isomap dimensionality
10
12
Fig. 4: ISOMAP Plot of variance (Eigen values) vs dimensionality:
Scree Plot Test
40
35
30
Eigenvalue
25
20
15
Knee Point = 12 PCs
10
5
0
0
10
20
30
40
50
60
Eigenvalue index - k
70
80
90
100
Fig. 5: PCA Plot of variance (Eigen values) vs dimensionality:
Table 5 comparison of PCA and ISOMAP Performance
NN Classify
M-Fold CV
PCA Analysis
Classified Misclassified
74%
26%
83%
17%
6.0 Conclusions
Data-based techniques are simple and cost
effective method of monitoring the health condition
of a system, as part of the predictive maintenance
strategy that seeks to improve and extend the
ISOMAP Analysis
Classified Misclassified
65%
35%
73.5%
26.5%
reliability and life of the system. ISOMAP and
PCA are employed to project the high-dimensional
data space to the lower dimensional subspace. The
low dimensional data set was analyzed to extract
changes in the feature for fault detection and
8
diagnosis. Data classification and visualization are
very effective means of discovering characteristics
or features encoded in a given data set. The GTE
data set was visualized in two-dimension using
scatter plot. The data-based model performance
evaluation results indicate that PCA is very suitable
and more effective in analyzing high-dimensional
data such as GTE dataset than ISOMAP, giving the
best classification for fault diagnosis. Thus PCA
data based technique for health condition
monitoring is an effective predictive maintenance
strategy which can easily extract unknown or
hidden features or geometrical structures of the
system parameter variables. These features can be
used to detect and diagnose system fault. The
weakness of ISOMAP in this project may be due to
the sparse nature of the GTE data set.
________________________________________________________________________
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
Chiang, L. H., E.L. Russell, and R.D. Braatz. (2001). Fault Detection and Diagnosis in Industrial
Systems. Springer-Verlag,
Eyoh, J. E., Eyoh, I. J., Umoh, U. A. and Udoh, E. N. (2011a), Health Monitoring of Gas Turbine
Engine using Principal Component Analysis Approach. Journal of Emerging Trends in
Engineering and Applied Sciences (JETEAS) 2 (4): 717-723
Eyoh, J. E., Eyoh, I. J., Umoh, U. A. and Umoeka, I. J. (2011b), Health Monitoring of Dimensional
Gas Turbine Engine (EngData) using ISOMAP Data-Based Analysis Approach. World Journal
of Applied Sciences and Technology (WOJAST) 3(2), 112-119.
Ghoshal, S., Roshan Shrestha, Anindya Ghoshal, Venkatsh Malepati, Somnath Deb, Krishna
pattipati and David Kleinman, (1999)“An Integrated Process For System Maintenance, Fault
Diagnosis and Support”, Invited Paper in Proc. IEEE Aerospace Conf., Aspen, Colorado.
Greitzer, F. L., Lars J. Kangas, Kristine M. Terrones, Melody A. Maynard, Bary W. Wilson,
Ronald A. Pawlowski, Daniel R. Sisk and Newton B. Brown, (1999). “Gas Turbine Engine
Health Monitoring and Prognostics”, Paper presented at the International Society of Logistics
(SOLE) 1999 Symposium, Las Vegas, Nevada, August 30 – September 2.
Isermann, R. (2006). “Fault-Diagnosis Systems – An Introduction from Fault Detection to Fault
Tolerance”, " Springer, Berlin.
Kadirkamanathan, V., (2008) “ACS 6304 – Health Care Monitoring”, Department of Automatic
Control & Systems Engineering, University of Sheffield, 21 – 25 January.
Martinez, W. L. and Angel R. Martinez, (2004) “Exploratory Data Analysis with MATLAB”,
(Computer Science and Data Analysis), Chapman & Hall/CRC, 2004 ...CRC Press.
Tenenbaum, J. B. (1998). “Mapping a Manifold of Perceptual Observations”, Department of Brain
and Cognitive Sciences, Massachusetts Institute of Technology, Cambridge, MA 02139.
Yang, P., Sui-sheng Liu, (2005)“Fault Diagnosis System for Turbo-Generator Set Based on Fuzzy
Network”, International Journal of Information Technology, Vol. 11 No. 12, 2005, pp. 76-84.
9
On the Probability Density Functions of Forster-Greer-Thorbecke (FGT) Poverty Indices
Osowole, O. I.+, Bamiduro, T.A+
Department of Statistics, University of Ibadan
dosowole@yahoo.com
Abstract
Distributional properties of poverty indices are generally unknown due to the fact that statistical inference for
poverty measures are mostly ignored in the field of poverty analysis where attention is usually based on
identification and aggregation problems. This study considers the possibility of using Pearson system of
distributions to approximate the probability density functions of Forster-Greer-Thorbecke (FGT) poverty
indices. The application of the Pearson system reveals the potentials of normal and four parameter distributions
in poverty analysis.
Keywords: Distributional properties, Pearson system of distributions, FGT poverty indices,
Normal distribution, Four parameter beta distribution.
_______________________________________________________________________________
1.0 Introduction
The poverty situation in Nigeria presents a
paradox, because despite the fact that the nation is
rich in natural resources, the people are poor. [1]
referred to this situation as poverty in the midst of
plenty. In 1992, for instance, 34.7 million Nigerians
(one-third of the population) were reported to be
poor, while 13.9 million people were extremely
poor [1]. The incidence of poverty increased from
28.1 percent in 1980 to 46.3 percent in 1985. The
poverty problem grew so worse in the 1990s that in
1996, about 65.6 percent of the population was
poor, while the rural areas accounted for 69.3
percent [2]. Recent data showed that in 2004, 54.4
percent of Nigerians were poor [3]. Also, more than
70 percent of the people are poor, living on less
than $1 a day. Similarly, Nigeria’s Human
Development Index (HDI) of 0.448 ranks 159th
among 177 nations in 2006, portraying the country
as one of the poorest in the world [4-5]. This
paradox was further highlighted in (Soludo, 2006).
He noted that Nigeria is a country abundantly
blessed with natural and human resources but the
potential remain largely untapped and even
mismanaged. With a population estimated at about
140 million, Nigeria is the largest country in Africa
and one sixth of the black population in the world. It
is the eight largest oil producers and has the sixth
largest deposit of natural gas in the world. The
growth in per capita income in the 1990s was zero
while the incidence of poverty in 1999 was 70%
[6].
Traditional approaches to measurement usually
start with the specification of poverty line and the
value of basic needs considered adequate for
meeting minimum levels of decent living in the
affected society. Poverty can also be measured
using the head count ratio which is based on the
ratio or percentage of the number of individuals or
households having incomes not equal to the poverty
line to the total number of individuals or
households [7-9]. Another method of measuring
intensity of poverty is the “income-gap” ratio. Here
the deviation of the incomes of the poor from the
poverty line is averaged and divided by the poverty
line [10]. These are the convectional approaches to
poverty analysis where the population is classified
into two dichotomous groups of poor and non-poor,
defined in relation to some chosen poverty line
based on household income/expenditure [11]. In the
last few years, poverty analyses made substantial
improvements by gradually moving from the
conventional
one-dimensional
approach
to
multidimensional approach [12-14].
10
Statistical inference for poverty and inequality
measures are widely ignored in the field of poverty
analysis where attention is usually based on
identification and aggregation problems [15] . The
implication of this is that distributional properties
of poverty and inequality indices are generally
unknown. This study therefore intends to
demonstrate how moments and cumulants of
Forster-Greer-Thorbecke (FGT) poverty indices
could be obtained from knowledge of their
probability density functions from the Pearson
system of distributions.
2.0
FORSTER-GREER
THORBECKE
(FGT)
POVERTY INDICES
In analyzing poverty, it has become customary to
use the so called FGT P-Alpha poverty measures
proposed by [11]. These FGT P-Alpha measures are
usually used to measure the poverty level. This is a
family of poverty indexes, based on a single
formula, capable of incorporating any degree of
concern about poverty through the “poverty
aversion” parameter, α. This measure is given as
1 n z − yi α
Pα = ∑1 (
) I ( z , yi )
(1)
n
z
where z is the poverty line; n is the total number of
is the
individuals in the reference population;
expenditure/income of the household in which
individual lives, α takes on values, 0, 1, and 2.
The quantity in parentheses is the proportionate
shortfall of expenditure/income below the poverty
line. This quantity is raised to a power α. By
increasing the value of α, the aversion to poverty as
measured by the index is also increased [16]. The
P-alpha measure of poverty becomes head count,
poverty gap and square poverty gap indices
respectively when α = 0, 1, and 2 in that order.
3.0
The Pearson System of Distributions
Several well known distributions like Gaussian,
Gamma, Beta and Student’s t -distributions belong
to the Pearson family. The
system
was
introduced by [17] who worked out a set of
four-parameter probability density functions as
solutions to the differential equation
f ′( x) P( x)
x−a
=
=
f ( x) Q( x) b0 + b1 x + b2 x 2
.(2)
where f is a density function and a , b 0 , b 1 and
b 2 are the parameters of the distribution. What
m a k e s the P e ar s o n’ s four-parameter s y s t e m
particularly
appealing i s the direct
c o r r e s p o n d e n c e between the parameters and
the central moments ( µ1 , µ 2 , µ3 and µ4 ) of the
distribution [18]. The parameters are defined as

µ ( µ + 3µ22 )
b1 = a = − 3 4

A

µ2 (4µ2 µ4 − 3µ32 ) 
b0 = −
(3)

A

(2 µ 2 µ4 − 3µ32 − 6 µ 23 ) 
b2 = −

A

The scaling parameter A is obtained from
A = 10 µ4 µ 2 − 18µ23 − 12 µ32
(4)
When the theoretical central moments are
replaced by their sample estimates, the above
equations define the moment estimators for the
Pearson parameters
a , b 0 ,b1
and b 2 . As
alternatives
to the basic four-parameter
systems, various extensions have been proposed
with the use of higher-order polynomials or
restrictions on the parameters. Typical extension
modifies (2) by setting P (x) = aO +a1 x so that
a0 + a1 x
f ′( x) P( x)
=
=
(5)
f ( x) Q( x) b0 + b1 x + b2 x 2
This parameterization characterizes t h e same
distributions b u t has the advantage t h a t a1 can
be zero and the values of the parameters are bound
when the fourth cumulant exists [19]. Several
attempts to parameterize t h e model using cubic
and quadratic curves have been made already by
Pearson and o t h e r s , but these systems proved
too cumbersome for general use. Instead the
simpler scheme with linear numerator and
quadratic d e n o m i n a t o r are more acceptable.
3.1
Classification
and
Selection
of
Distributions in the Pearson System
There are different ways to classify the
distributions generated by the roots of the
polynomials in (2) and (5). Pearson himself
organized the solution to his equation i n a
system of twelve classes identified by a number.
The numbering c r i t e r i o n has no systematic
11
b a s i s and it has varied depending on the
source. An alternative approach suggested
by [20] for distribution selection based on
two statistics t h a t are functions of the four
Pearson parameters will be adopted. The scheme
is presented i n Tables 1 and 2 where D and λ
denote the selection criteria. D and λ are defined
as
D = b0b2 − b12 


b12
λ=

b0b2

(6)
Table 1: Pearson Distributions
The table provides a classification of the Pearson Distributions, f(x) satisfying the differential equation
(a + a x)
b2
df
P( x)
2
0 1
1 , are
.
The
signs
and
values
for
selection
criteria,
D
=
b
b
−
b
and
λ
=
(1)
=
=
0 2
1
f dx Q ( x ) (b +b x+b x2 )
bb
02
0 1 2
given in columns three and four.
Table 1: Person Distributions
P(x) = a0 , Q(x) = 1
1.
Restrictions
a0 < 0
Support
R+
D λ
0 0/0
Density
γ e −γ x , γ > 0
P(x) = a0 , Q( x) = b2 x( x + α )
Restrictions
2(a). α > 0
2(b)
.
α>0
D
λ
<0 ∞
Support
[ - α , 0]
∞
[ - α , 0]
<0
Density
m +1
( x + α ) m , m < −1
m +1
α
m +1
α
P(x) = a0 , Q( x) = b0 + 2b1 x + x = ( x − α )( x − β ), α < β
m +1
( x + α ) m , −1 < m < 0
2
3(a).
Restrictions
a0 ≠ 0
D
λ
< 0 >1
Support
[ β, ∞]
< 0 >1
[ -∞,α]
0 <α < β
3(b)
.
4.
a0 ≠ 0
α <β <0
a0 ≠ 0
< 0 < 0 [ α, β]
α <0<β
5.
Density
( β − α ) − ( m + n +1)
( x − α )m ( x − β ) n
B(−m − n − 1, n + 1)
m > −1, n > −1, m ≠ 0, n ≠ 0, m = − n
( β − α ) − ( m+ n +1)
( x − α )m ( x − β )n
B(−m − n − 1, m + 1)
m > −1, n > −1, m ≠ 0, n ≠ 0, m = − n
P( x) = a0 + a1 x, Q( x) = 1
Restrictions D λ
0 0/0
a1 ≠ 0
α 2m β 2n
( x − α )m ( x − β )n
(α + β )
B(m + 1, n + 1)
m > −1, n > −1, m ≠ 0, n ≠ 0, m = −n
m+ n+1
Support
R
Density
−
1
e
2πσ
( x − µ )2
2σ 2
12
Table 2: Pearson Distributions (Continued)
P( x) = a0 + a1 x, Q( x) = x − α
Restrictions D
λ
Support
Density
6.
<
0
[α,∞]
∞
a1 ≠ 0
k m +1
( x − α ) − m e− k ( x −α ) , k > 0
Γ(m + 1)
P ( x) = a0 + a1 x, Q ( x) = b0 + 2b1 x + x 2 = ( x − α )( x − β ), α ≠ β
Restrictions D
λ
Support
Density
7(a) a1 ≠ 0
< 0 >1 [ β, ∞]
( β − α ) − ( m + n +1)
( x − α )m ( x − β ) n
.
0 <α < β
B (−m − n − 1, n + 1)
7(b)
.
8.
a1 ≠ 0
<0
>1
[ -∞,α]
<0
< 0 [ α, β]
α <β <0
a1 ≠ 0
α <0<β
m > −1, n > −1, m ≠ 0, n ≠ 0, m ≠ − n
( β − α ) − ( m+ n +1)
( x − α )m ( x − β )n
B (−m − n − 1, m + 1)
m > −1, n > −1, m ≠ 0, n ≠ 0, m ≠ − n
α 2m β 2n
( x − α )m ( x − β )n
B (m + 1, n + 1)
(α + β )
m > −1, n > −1, m ≠ 0, n ≠ 0, m ≠ −n
m + n +1
9.
P ( x) = a0 + a1 x, Q ( x) = b0 + 2b1 x + x 2 = ( x − α )( x − β ), α = β
0 1 [ α, ∞]
a1 > 0
γ m−1
−γ
( x − α ) − m e x , γ > 0, m > 1
α =β
Γ(m − 1)
10.
P ( x) = a0 + a1 x, Q ( x) = b0 + 2b1 x + x 2 , complex roots
Restrictions D
λ
Support Density
R
a0 = 0, a1 < 0 >0 0
α 2 m −1
1
( x 2 + β 2 )− m , m >
2
1
1
2
b1 = 0, b0 = β
B(m − , )
2 2
β ≠0
11.
a0 ≠ 0, a1 < 0
b1 ≠
a0
a1
>0
0>
<1
R
c(b0 + 2b1 x + x 2 )− m e
− var c tan(
( x +b1 )
β)
1
m > , β = b0 − b12
2
The advantage of this approach in statistical
modeling in the P ea rs on framework i s its
simplicity. Implementation is done in accordance
with the following steps:
(4) Select an appropriate distribution
from Tables (1) and (2) based on t h e
signs of the values of the
selection criteria.
(1) Estimate m o m e n t s from data.
(2) Calculate t h e Pearson parameters a ,
b 0 , b 1 and b 2 using (3) and (4).
(3) Use the e s t i m a t e s o f t h e
parameters
to compute the selection
criteria
D and λ as given in (6).
4.0
Bootstrapping
Poverty indices are complex in nature and this
makes direct analytic solutions very tedious and
complex. Alternative numerical solutions are
possible through simulation Bootstrapping.
Bootstrapping is essentially a re-sampling method.
That is, re-sampling is a Monte-Carlo method of
13
simulating data set from an existing data set,
without any assumption on the underlying
population. Bootstrapping was invented by [21-22]
and further developed by [23]. It is based on resampling with replacement from the original
sample. Thus each bootstrap sample is an
independent random sample of size n from the
empirical distribution. The elements of bootstrap
samples are the same as those of the original data
set. Bootstrapping, like other asymptotic methods,
is an approximate method, which attempts to get
results for small samples (unlike other asymptotic
methods). The estimates of the parameters of the
selection criteria for the purpose of selecting
appropriate probability distributions from the
Pearson system for head count, poverty gap and
square poverty gap indices were obtained through
bootstrap simulation method. The bootstrap sample
size was 10,000 and the number of iterations was
5,000.
5.0
Results and Discussion
The methods presented are applied to The
Nigerian Living Standard Survey (NLSS, 2004)
data. The survey was designed to give estimates at
National, Zonal and State levels. The first stage was
a cluster of housing units called numeration Area
(EA), while the second stage was the housing units.
One hundred and twenty EAs were selected and
sensitized in each state, while sixty were selected in
the Federal Capital Territory. Ten EAs with five
housing units were studied per month. Thus a total
of fifty housing units were canvassed per month in
each state and twenty-five in Abuja. Data were
collected on the following key elements:
demographic characteristics, educational skill and
training, employment and time use, housing and
housing conditions, social capital, agriculture,
income, consumption expenditure and non-farm
enterprise. The total number of households in the
survey was 19,158.
The estimates of the selection criteria for the
selection of probability distributions from the
Pearson system were obtained as shown in Table 3.
Based on the values and signs of these criteria, the
normal and four parameter beta distributions were
selected for the poverty indices based on the
classifications in Tables 1 and 2. The normal
distribution was selected for the head count index
while, the four parameter beta distribution was
selected for both poverty gap and square poverty
gap indices respectively. The estimates of the
parameters of these selected distributions were
equally estimated as shown in Tables 4, 5 and 6.
Table 3: Estimates of Selection Criteria
b2
( D = b0b2 − b12 and λ = 1 ) for FGT Poverty
b0b2
Indices
b0
b1= a
b2
D
λ
A
P0
Poverty Head
Count Index
-1.13687 X 10-5
-4.91928 X 10-5
-1.45081 X 10-2
1.62518 X 10-7
3.08817 X 10-6
2.17318 X10-14
P1
Poverty Gap
Index
-3.40804 X 10-6
-3.84213 X 10-5
6.80230 X 10-3
-2.46587 X 10-8
6.36771 X 10-2
4.32546 X 10-16
P2
Square Poverty
Gap Index
-1.62081 X 10-6
-1.33289 X 10-5
6.40434 X 10-3
-1.05579 X 10-8
-1.71151 X 10-2
4.67674 X 10-17
Table 4: Parameter Estimates of the Normal
Distribution for Head Count Poverty Index
Parameter Estimate
µ
0.52096
σ
0.00345
Table 5: Parameter Estimates of the Four
Parameter Beta Distribution for Poverty Gap
Index
Parameter Estimate
α1
224.73
α2
388.02
a
0.17752
b
0.27147
Table 6: Parameter Estimates of the Four Parameter
Beta Distribution for Square Poverty Gap Index
Parameter
α1
α2
a
b
Estimate
47.953
48.085
0.10164
0.12648
14
indices. The normality confirms the applicability of
6.0
Conclusion
laws of large numbers and the consequent validity
The probability distributions of head count,
of the central limit theorem. Hence, study on
poverty gap and square poverty gap indices have
poverty indices should involve large samples. The
been determined. The distributions appropriate for
selection of the beta distribution for the two indices
the indices obtained using the procedure given by
may be due to the fact that the Beta distribution is
Andreev for the selection of probability
often used to mimic other distributions when a
distributions from the Pearson system of
vector of random variables is suitably transformed
distributions were the normal distribution for head
and normalized.
count index and the four parameter beta distribution
for both poverty gap and square poverty gap
_______________________________________________________________________________
References
[1].
[2].
[3].
[4].
[5].
[6].
[7].
[8].
[9].
[10].
[11].
[12].
[13].
[14].
[15].
[16].
[17].
[18].
[19].
World Bank (1996) “Poverty in the Midst of Plenty: The challenge of growth
with
inclusion in Nigeria” A World Bank Poverty Assessment, May 31, World Bank, Washington,
D.C.
Federal Office of Statistic (FOS) (1999), Poverty and Agricultural sector in
Nigeria
FOS, Abuja, Nigeria.
Federal Republic of Nigeria (FRN) (2006). Poverty Profile for Nigeria. National Bureau
of Statistics (NBS) FRN.
United Nations Development Program (UNDP) (2006). Beyond scarcity: Power, poverty
and the global water crisis. Human Development Report 2006.
IMF (2005). Nigeria: Poverty Reduction Strategy Paper— National Economic
Empowerment and Development Strategy. IMF Country Report No. 05/433.
Soludo (2006) ‘’Potential Impacts of the New Global Financial Architecture
on Poor
Countries’’: Edited by Charles Soludo, Monsuru Rao, ISBN
9782869781580,
2006,
CODERSIA, Senegal, paperback. 80 pgs.
Bardhan, P. K. (1973) “On the Incidence of Poverty in Rural India”. Economic and
Political Weekly,. March
Ahluwalia, M.S. (1978) “Inequality, Poverty and Development”. Macmillan
Press
U.K.
Ginneken, W. V.(1980), “Some Methods of Poverty Analysis: An Application to Iranian
Date,” World Development, Vol. 8
World Bank (1980), Poverty and Basic Needs Development Policy Staff Paper,
Washington D.C.
Foster, James, J. Greer and Eric Thorbecke.(1984) . “A Class of Decomposable Poverty
Measures,”Econometrica, 52(3): 761-765.
Hagenaars A.J.M. (1986), The Perception of Poverty, North Holland, Amsterdam.
Dagum C. (1989), “Poverty as Perceived by the Leyden Evaluation Project. A Survey of
Hagenaars’ Contribution on the Perception of Poverty”, Economic Notes, 1, 99-110.
Sen A.K. (1992), Inequality Reexamined, Harvard University Press, Cambridge (MA).
Sen, A. (1976) “An Ordinal Approach to Measurement”, Econometrica, 44, 219- 232.
Boateng, E.O., Ewusi, K., Kanbur, R., and McKay, A. 1990. A Poverty Profile for Ghana,
1987-1988 in Social Dimensions of Adjustment in Sub-Saharan Africa, Working Paper 5. The
World Bank: Washington D.C.
Pearson, K.1895. Memoir on Skew Variation in Homogeneous Material.
Philosophical Transactions of the Royal Society. A186: 323-414
Stuart, A. and Ord. J.1994. Kendall’s Advanced Theory of Statistics, Vol. I:
Distribution Theory. London: Edward Arnold.
Karvanen, J.2002. Adaptive Methods for Score Function Modeling in Blind Source
15
Separation. Unpublished Ph.D. Thesis, Helsinki University of Technology.
[20]. Andreev, A., Kanto, A., and Malo, P.2005. Simple Approach for Distribution Selection in
the Pearson System. Helsinki School of Economics Working Papers: W-388.
[21]. Efron, B.1982. The Jacknife, the Bootstrap, and Other Resampling Plans. Philadelphia:
SIAM.
[22]. Efron, B. 1983. Bootstrap Methods; Another Look at the Jacknife. The Annals of Statistics.
7: 1-26.
[23]. Efron, B. and R.J. Tibshirani.1993. An Introduction to the Bootstrap. London: Chapman
and Hall.
16
Comparison of Three Criteria for Discriminant Analysis Procedure
Nwosu, Felix D., Onuoha, Desmond O. and Eke Charles N.
Department of Mathematics and Statistics
Federal Polytechnic Nekede, Owerri, Imo State.
E-mail: desonuoha@yahoo.com
Abstract
This paper presents a fisher’s criterion, Welch’s criterion, and Bayes criterion for performing a discriminant
analysis. These criteria estimates a linear discriminant analysis on two groups (or regions) of contrived
observations. The discriminant functions and classification rules for these criteria are also discussed. A linear
discriminant analysis is performed in order to determine the best criteria among Fisher’s criterion, Welch’s
criterion and Bayes criterion by comparing their apparent error rate (APER). Any of these criteria with the
least error rate is assumed to be the best criterion. After comparing their apparent error rate (APER), we
observed that, the three criteria have the same confusion matrix and the same apparent error rate. Therefore
we conclude that none of the three criteria is better than each other.
Kay Words: Fisher’s criterion, Welch’s criterion, Bayes criterion and Apparent Error rate
___________________________________________________________________________________
1.
Introduction:
Discriminant Analysis is concerned with the
problem of classification. This problem of
classification arises when an investigator makes a
number of measurements on an individual and
wishes to classify the individual into one of several
categories or population groups on the basis of
these measurements. This implies that the basic
problem of discriminant analysis is to assign an
observation X, of more distinct groups on the basis
of the value of the observation. In some problems,
fairly complete information is available about the
distribution of X in the two groups. In this case we
may use this information and treat the problem as if
the distributions are known. In most cases, however
information about the distribution of X comes from
a relatively small sample from the groups and
therefore, slightly different procedures are used.
The Objectives of Discriminant Analysis
includes: To classify cases into groups using a
discriminant prediction equation; to test theory by
observing whether cases are classified as predicted;
to investigate differences between or among
groups; to determine the most parsimonious way to
distinguish among groups; to determine the percent
of variance in the dependent variable explained by
the independents; to assess the relative importance
of the independent variables in classifying the
dependent variable and to discard variables which
has little relevance in relation to group distinctions.
In this study, we wish to determine the best
criterion among the three criteria namely; Fisher’s
criterion, Welch’s criterion and Bayes criterion for
good discriminant functions, by comparing their
apparent error rate (APER) while the significance is
for detecting the variables that allow the researcher
to discriminate between different groups and for
classifying cases into different groups with a better
than chance accuracy.
2.
Related Literature:
Anderson[1]
viewed
the
problem
of
classification as a problem of “statistical decision
functions”. We have a number of hypotheses which
proposes is that the distribution of the observation
is a given one. If only two populations are admitted,
we have an elementary problem of testing one
hypothesis of a specified distribution against
another.
Lachenbruch (1975) viewed the problem of
discriminant analysis as that of assigning an
unknown observation to a group with a low error
rate. The function or functions used for the
assignment may be identical to those used in the
multivariate analysis of variance.
Johnson
and
Wichern
(1992)
defined
discriminant analysis and classification as
multivariate techniques concerned with separately
distinct set of observations (or objects) and with
17
allocating new observation (or object) to previously
defined groups. They defined two goals namely:
Goal 1: To describe either graphically (in at most
three dimensions) or algebraically the differential
features of objects (or observations) from several
known collections (or populations) and Goal 2: To
sort observations (or objects) into two or more
labeled classes. The emphasis is on deriving a rule
that can be used to optimally assign a new
observation to the labeled classes. They used the
term discrimination to refer to Goal 1 and used the
term classification or allocation to refer to Goal 2.
A more descriptive term for goal 1 is separation.
They also explained that a function that separates
may sometimes serve as an allocator or
classificatory and conversely an allocation rule may
suggest a discriminator procedure. Also that goal 1
and 2 frequently overlap and the distinction
between separation and allocation becomes blurred.
According to Bartlett M.S. (1951); Discriminant
function analysis is used to determine which
variables discriminate between two or more
naturally occurring groups.
Costanza W.J. and Afifi A.A. (1979)
computationally stated that discriminant function
analysis is very similar to analysis of variance
(ANOVA).
Theoretical Basis by Lachenbruch P.A. [6]
elaborated that the basic problem in discriminant
analysis is to assign an unknown subject to one of
two or more groups on the basis of a multivariate
observation. It is important to consider the costs of
assignment, the a priori probabilities of belonging
to one of the groups and the number of groups
involved. The allocation rule is selected to optimize
some function of the costs of making an error and
the a priori probabilities of belonging to one of the
groups. Then the problem of minimizing the cost of
assignment is to minimize the following equation
Min ∑ ∑ P (Dj/∏i) Pi Cji
3.0
The Criterion:
3.1
Fishers Criterion:
Fisher (1936) suggested using a linear
combination of the observations and choosing the
coefficients so that the ratio of the differences of
the means of the linear combination in the two
groups to its variance is maximized. For classifying
observation into one of two population groups,
fisher considered the linear discriminant function
у=λ1X. Let the mean of у in population I (∏1) be
λ1µ1, and the mean of у in ∏2 be λ1µ2, its variance
is λ1∑λ in either population where ∑ = ∑1 = ∑2.
Then he chooses λ to
(λ µ
Maximize Φ =
1
− λ1 µ 2
λ1Σλ
1
)
2
(3.1.1)
Differentiating (3.1.1) with respect to λ, we have
dΦ 2(λ1µ1 − λ1µ2 )(µ1 − µ2 )λ1Σλ − 2λΣ(λ1µ1 − λ1µ2 )2
=
2
dλ
λ1Σλ
(
)
(3.1.2)
Equating (3.1.2) to zero, we have
2(λ1µ1– λ1µ2)(µ1– µ2) λ1Σλ = 2λΣ (λ1µ1 – λ1µ2)2
Σλ (λ1 µ1 − λ1 µ 2 )
(3.1.3)
µ1–µ =
λ1Σλ
Since λ is used only to separate the populations, we
may multiply λ by any constant we desire. Thus λ is
proportional to Σ −1 ( µ1 − µ 2 ). The assignment
procedure is to assign an individual to ∏1, If Y =
(µ1– µ2)1 Σ-1X is closer to Y1 = (µ1– µ2)1 Σ-1µ1 than
to Y2 = (µ1– µ2)1Σ-1µ2 and an individual is assigned
to Π2 if Y = (µ1– µ2)1 Σ-1 X is closer to Y2 = (µ1–
µ2)1 Σ-1 µ1 than to Y1 . Then midpoint of the interval
between Y1 and Y2 is
y1 + y 2
= ½ (µ1 + µ2)1 Σ-1 (µ1 – µ2).
2
This is used as the cut off point for the assignment.
The difference between Y1 and
Y2 is
Y1 – Y2 = (µ1 – µ2) 1∑ -1µ 1 (µ1 – µ2)1∑-1 µ 2 = (µ1 –
µ2)1∑- 1(µ1 – µ2) = δ2
δ2 is called the Mahalanobi’s (squared) distance for
known parameters. If the parameters are not known,
it is the usual practice to estimate them by X 1 , X 2
and S where X 1 is the mean of a sample from ∏1,
X 2 is the mean of a sample from ∏2 and S is the
pooled sample variance-covariance matrix from the
two groups. The assignment procedure is to assign
18
an individual to ∏1 if Y = ( X 1 – X 2 )1S-1X is closer
X 2 )1S-1 X 1 than to Y2 = ( X 1 –
to Y1 = ( X 1 –
X 2 )1S-1 X 2
while an individual is assigned to ∏2 if Y = ( X 1 –
X 2 )1S-1X is closer to Y2 = ( X 1 − X 2 )1 S −1 X 2 than
to Y1 = ( X 1 – X 2 )1S-1 X 1
The midpoint of the interval between Y1 and Y2 is
y1 + y 2
= ½ ( X 1 + X 2 )1S-1( X 1 – X 2 ).
2
Y is closer to Y1 if |Y – Y1 | < |Y – Y2 | which
occurs if Y > ½ ( Y1 + Y2 ) since Y1 > Y2 , The
difference between Y1 and Y2 is Y1 – Y2 = ( X 1 –
X 2 )1S-1 X 1 – ( X 1 – X 2 )1S-1 X 2
= ( X 1 – X 2 )1S-1( X 1 – X 2 ) = D2 which is called
the Mahalanobis (squared) distance for unknown
parameters.
The distribution of D2 can be used to test if there
are significant differences between the two groups
(or Regions). We consider the two independent
random samples (Xij, j = 1, 2, . . . n1) and (X2j, j = 1,
2, . . . n2) from Nk(µ1, ∑) and Nk(µ2, ∑)
respectively. We test the hypothesis that both
samples came from the same normal distribution,
that is,
H0: µ 1= µ 2 versus H1: µ 1≠ µ 2.
∑ (x
)(
n1
Let
A1
=
j =1
∑ (x
n2
j =1
2j
)(
− x 2 x2 j − x 2
)
1
− x1 x1 j − x1 ;
1j
A2
=
)
1
The
pooled
estimator
S
of
∑
is
A1 + A2
n1 − 1)S1 + (n2 − 1)S 2
(
S=
=
and
is
n1 + n 2 − 2
n1 + n 2 − 2
unbiased for ∑
It is the property of the Wishart Distribution that if
Xij ~ iidNk(N, ∑) 1 < j < n then
∑ (x
n
A =
j =1
ij
)(
)
1
− x1 xij − x 1 ~ Wk (∑, n-1); therefore
(n1 + n2 − 2)S
independent
~
of
Wk(∑,
( X1 –
(n1 + n2 − 2)
X2 )
and
is
which
is
 1
1  
N k  0,  + Σ  when the null hypothesis is
  n1 n 2  
true.
x1 + x 2
~ N k (0, Σ ) and is Independent of S.
1
1
+
n1 n2
Therefore, T2 = V X1 D-1X, V > k is the Hotelling’s
T2 based on V degrees of freedoms where X and D
are independent.
Here we have X = X 1 – X 2 and D = (n1 + n2 − 2 )S ;
−1
(
)
(
)
1
1 1
−1
T =  +  x1 − x2 ((n1 + n2 − 2)S) x1 − x2 (n1 + n2 − 2)
 n1 n2 
2
(
) (
)
1
n1n2
x1 − x 2 S −1 x1 − x 2
n1 + n2
(3.1.4)
If X2 ~ Nk(0, ∑) and D ~Wk (∑, V), D, X is
kv
independent, then T 2 ~
Fk , n − k + 1
n − k +1
Therefore,
1
n1n2
T2 =
x1 − x2 S −1 x1 − x 2
n1 + n2
kv
~
Fk , n − k + 1
n − k +1
F=
1
n1 n2 (n1 + n2 − k − 1)
x1 − x 2 S −1 x1 − x 2
n1 + n2 (n1 + n2 − 2)k
n n (n + n2 − k − 1) 2
D
F= 1 2 1
(3.1.5)
n1 + n2 (n1 + n2 − 2)k
n n (n + n2 − k − 1) 2
The variable F = 1 2 1
D
n1 + n2 (n1 + n2 − 2)k
=
(
(
) (
) (
)
)
where n1 and n2 are the sample sizes in ∏1 and ∏2
respectively and K is the number of variables, has
an F-distribution with F and n1 + n2 − k − 1 degrees
y1 + y 2
as a cut off point
2
can be improved upon if the apriori probabilities of
∏1 and ∏2 are not equal.
of freedom. The use of
19
3.2
Welch’s Criterion
An alternative way to determine the discriminant
function is due to Welch (1939). Let the density
functions of ∏1 and ∏2 be denoted by F1(X) and
F2(X) respectively. Let q1 be the proportion of ∏1 in
the population and q2 = (1 – q1) be the proportion of
∏2 in the population. Suppose we assign X to ∏1 if
X is in some region R1 and to ∏2 if X is in some
region R2. We assume that R1 and R2 are mutually
exclusive and their union includes the entire space
R. The total probability of misclassification is,
T ( R, F ) = q1 ∫ f 1 ( x)dx + q 2 ∫ f 2 ( x)dx
R
R
= q1 (1 − ∫ f1 ( x)dx) + q 2 ∫ f 2 ( x)dx
R
R
= q1 + ∫ (q 2 f 2 ( x) − q1 f 1 ( x))dx
(3.2.1)
R
This quantity is minimized if R1 is chosen such that
q2f2(x) = q1f1(x) < O for all points in R1
Thus the classification rule is:
f ( x) q 2
Assign X to ∏1 if 1
>
(3.2.2)
f 2 ( x) q1
And to ∏2 if otherwise; it is pertinent to note
that this rule minimizes the total probability of
misclassification.
An important special case is when ∏1 and ∏2 are
multivariate normal with
means µ1 and µ2 and common covariance matrix ∑.
The density in population ∏1 is
1
f i ( x) =
exp{−1/ 2( x − µi )1 Σ1 ( x − µi )}
p/2
1/ 2
(2π ) | Σ |
(3.2.3)
The quantity on the left of equation 3.2.5 is called
the true discriminant function DT(X).
Its sample analogue is
[
)] (
(
)
1
DT ( X ) = X − 1 X 1 + X 2 S −1 X 1 − X 2 (3.2.6)
2
The coefficient of X is seen to be identical with
Fishers result for the linear discriminant function.
The function DT(X) is a linear transformation of X
and knowing its distribution will make it possible to
calculate the error rates that will occur if DT(x) is
used to assign observation to ∏1 and ∏2. Since X is
multivariate normal, DT(x) being a linear
combination of X is normal. The means of DT(x) if
X comes form ∏1 is
1
D ( x) 
E  T
= µ1 − 1 (µ1 + µ 2 ) Σ −1 (µ1 − µ 2 )

Π1 
2

[
[
]
]
= µ 2 − 1 µ1 − 1 µ 2 Σ −1 (µ1 − µ 2 )
2
2
1
= − 1 (µ1 − µ 2 ) Σ −1 (µ1 − µ 2 )
2
= 1 δ2
2
1
Where δ2 = (µ1 - µ2)1∑-1 (µ1 - µ2)
In ∏2, the mean of DT(x) is
1
D ( x) 
−1
E  T
 = µ 1 − 1 2 (µ 1 + µ 2 ) Σ (µ 1 − µ 2 )
Π
1

1
= µ 2 − 1 µ1 − 1 µ 2 Σ −1 (µ1 − µ 2 )
2
2
1
= − 1 (µ1 − µ 2 ) Σ −1 (µ1 − µ 2 )
2
= 1 δ2
2
[
[
]
]
The ratio of the densities is
In either population the variance is
1
1
EDT (x) −DT (µi )2 = E(µ1 −µ2 ) ∑−1(x −µi )(x −µi ) ∑−1(µ1 −µ2 )
f1 ( x) exp{−1 / 2( x − µ1 ) Σ ( x − µ1 )}
=
f 2 ( x) exp{−1 / 2( x − µ 2 )1 Σ1 ( x − µ 2 )}
1
1
= (µ1 − µ 2 ) ∑ −1 E ( x − µ i )( x − µ i ) ∑ −1 (µ1 − µ 2 )
1
=
exp[−1/ 2{(x − µ1 )1 Σ1 (x − µ1 ) − (x − µ2 )1 Σ1 (x − µ2 )}]
=exp−1 −x1Σ−1µ1 −µ1Σ−1x1 +µ11Σ−1µ1 +x1Σ−1µ2 +µ2Σ−1x1 −µ12Σ−1µ2
2
1
= exp X − 1 (µ1 + µ 2 ) Σ −1 (µ1 − µ 2 ) (3.2.4)
2
The optimal rule is to assign the unit X to ∏1 if
1
q
DT ( X ) = X − 1 (µ1 + µ2 ) Σ−1 (µ1 − µ2 ) > in 2
=
2
q1
(3.2.5)
{ [
{[
}
]
[
]
]}
1
= (µ1 − µ 2 ) ∑ −1 (µ1 − µ 2 ) = δ 2
1
The quantity δ2 is the population Mahalanobis
(squared) distance.
3.3
Bayes Criterion
A Bayesian criterion for classification is one that
assigns an observation to a population with the
greatest posterior probability. A Bayesian criterion
20
for classification is to place the observation in ∏1 if
p (∏1/x) > p (∏2/x).
Π
By Bayes theorem p i  = p (Π : x)
p( x)
 x
q1 f i ( x)
=
;
q1 f 1 ( x) + q 2 f 2 ( x)
Hence the observation X is assigned to ∏1 if
q1 f 1 ( x)
q 2 f 2 ( x)
>
(3.3.1)
q1 f 1 ( x) + q 2 f 2 ( x) q1 f 1 ( x) + q 2 f 2 ( x)
The above rule reduces to assigning the observation
f ( x) q 2
to ∏1 if 1
>
(3.3.2)
f 2 ( x) q1
and to ∏2 otherwise.
Where
1
f1 ( x)
= exp X − 1 (µ1 + µ 2 ) Σ −1 (µ1 − µ 2 )
2
f 2 ( x)
q1 = the proportion of ∏1 in the population.
q2= (1-q1) = the proportion of ∏2 in the population.
{[
}
]
3.4
Probabilities of Misclassification
In constructing a procedure of classification, it is
desired to minimize the probability of
misclassification or more specifically, it is desired
to minimize on the average the bad effects of
misclassification. Suppose we have an observation
from either population ∏1 or population ∏2 the
classification of the observation depends on the
vector of measurements.
X1= (X1, X2, . . . , Xk)1 on the observation. We set
up a rule that if an observation is characterized by
Population
Π1
Π2
certain sets of values of X1, X2,. . .,Xk, we classify it
as from ∏1, if it has other values, we classify it as
from ∏2. We think of an observation as a point in a
K-dimensional space. We divide the space into two
regions or groups. If the observation falls in R1, we
classify it as coming from population ∏1, and if it
falls in R2, we classify it as coming from population
∏2.
In following a given classification procedure, the
statistician can make two kinds of errors in
classification. If the observation is actually from
∏1, the statistician or researcher can classify it as
coming from ∏2; or if it is from ∏2, the statistician
may classify it as from ∏1. We need to know the
relative undesirability of these two kinds of
misclassification.
Let the probability that an observation comes
from population ∏1 be q1 and from population ∏2
be q2. Let the density function of population ∏1 be
f1(x) and that of population ∏2 be f2(x). Let the
regions of classification from ∏1 be R1 and from ∏2
be R2. Then the probability of correctly classifying
an observation that is actually drawn from ∏1 is
∫ f ( x)dx where dx = dx1, dx2, . . , dxk and the
R1
probability of misclassifying an observation from
∏1 is P1 = ∫ f ( x)dx
R2
Similarly the probability of correctly classifying an
observation from ∏2 is ∫f2(x) dx and the probability
of misclassifying such an observation is
P2 = ∫ f ( x)dx ; then the total probability of
R2
misclassification
T ( R; f ) = q1 ∫ f 1 ( x)dx + q 2 ∫ f 2 ( x)dx (3.4.1)
R2
is
R1
Table 3.4.1: confusion matrix
Statisticians’ decision
Π1
Π2
Correct Classification
P1
P2
Correct Classification
Probabilities of misclassification can be
computed for the discriminant function. Two cases
have been considered.
(i)
When the population parameter are know.
(ii)
When the population parameter are not
known but estimated from samples drawn from the
two populations.
21
3.5
Apparent Error Rates (APER)
One of the objectives of evaluating a
discriminant function is to determine its
performance in the classification of future
observations.
When
the
(APER)
is
T ( R; f ) = q1 ∫ f 1 ( x)dx + q 2 ∫ f 2 ( x)dx
R2
R1
If f1(x) is multivariate normal with mean µ1 and
covariance ∑, we can easily calculate these rates.
When the parameters are not known a number of
error rates may be defined. The function T(R,F)
defines the error rates (APER). The first argument
is the presumed distribution of the observation that
will be classified.
4.0
Data Analysis
Consider to carry out a linear discriminant analysis on two groups (or regions) of contrived
observations.
A
B
X1
6
7
9
8
8
10
X1
11
15 22
X2
7
5
10
8
9
9
X2
13
16 20
17
16
4.1: Using Fishers Criterion
For A
 Σx12 − Nµ12
A = 
 Σx 2 x1 − Nµ 2 µ1
Σx1 x 2 − Nµ1 µ 2 

Σx 22 − Nµ 22 
10 9 

= A = 
 9 16 
For A
ΣX 1 = 48, ΣX 2 = 48, ΣX 12 = 394, ΣX 22 = 400, ΣX 1 X 2 = 393, X 1 = µ1 = 8, X 2 = µ 2 = 8, N = 6
For B
ΣX 1 = 90, ΣX 2 = 90, ΣX 12 = 1432, ΣX 22 = 1398, ΣX 1 X 2 = 1409, X 1 = µ1 = 15, X 2 = µ 2 = 15, N = 6
 Σx 2 − N x12
1
B=
 Σx x − N x 2 x1
 2 1
S=
A+ B
n1 + n 2 − 2
(
Σx1 x 2 − N x 1 x 2 
2
Σx 22 − N x 2 
=
=>
 82 59 

B = 
59
48


 9.2 6.8 


 6.8 6.4 
)
1
Y = x1 − x 2 S −1 X
Y = 0.2212X1 - 1.3293X2 which is the discriminant function.
(
)
1
Y1 = x1 − x 2 S −1 X 1
=>
8
Y1 = (0.2212 – 1.3293)   = -8.8648
8
15 
Y2 = (0.2212 – 1.3293)   = -16.6215
15 
Cut off point =
Y1 + Y2
and this is also referred to as the mid point and it’s equal to -12.74315.
2
22
12 13
11 14
Assignment procedure:
Assign observation with measurement X to ∏1 if Y >
Y1 + Y2
Y + Y2
and assign to ∏2 if Y ≤ 1
2
2
Discriminant scores
Y = 0.2212X1 -1.3293X2
A
B
-7.9779
-14.8477
-5.0981
-17.9508
-11.3022
-21.7196
For Group A
-7.9779 – (-12.74315) = 4.7653 > ∏1
-5.0981 – (-12.74315) = 7.6451 > ∏1
-11.3022 – (-12.74315)= 1.4410 > ∏1
-8.8648 – (-12.74315) = 2.5491 > ∏1
-10.1941 – (-12.74315)= 2.5491 > ∏1
-9.7517 – (-12.74315) = 2.9915 > ∏1
Population ∏1
Population ∏2
-8.8648
-18.8356
-10.1941
-17.5084
-9.7517
-15.7346
For Group B
-14.8477 – (-12.74315) = -2.1046 < ∏2
-16.9508 – (-12.74315) = -5.2077 < ∏2
-21.7196 – (-12.74315) = -6.0925 < ∏2
-18.8356- (-12.74315) = -6.0925 < ∏2
-17.5084 – (-12.74315) = 0.7753 > ∏1
-15.7346 – (-12.74315) = -2.9915 < ∏2
Tables 4.1.1. Confusion matrix
Statistician decision
∏1
∏2
6
0
1
5
The probability of misclassification
P(2/1) = 0/6 = 0
P (1/2) = 1/6
Apparent error rate (APER)
2 1
n  + n 
1
1
1 2
Error rate =
= ; hence the APER =
TotalΠ 1 + TotalΠ 2 12
12
4.2 : Using Welch’s Criterion
The classification rule is:
f ( x) q 2
Assign X to ∏1 if 1
>
and to ∏2 if otherwise.
f 2 ( x) q1
1
f1 ( x)
q
= exp X − 1 (µ1 + µ 2 ) Σ −1 (µ1 − µ 2 ) > 2
2
f 2 ( x)
q1
{[
}
]
Taking the Lim of both sides; we have
{[X − 12 (µ + µ )] Σ
1
1
2
−1
}> In qq
(µ 1 − µ 2 )
{[X − 12 (µ + µ )] Σ
1
1
2
−1
} > In qq
(µ 1 − µ 2 )
2
therefore;
1
2
1
Where
23
DT(x) =
DT(x) is called the true discriminant function and q1 = q2 since they have equal sample size.
The optimal rule is to assign the unit X to ∏1 if
1
q
DT(x) = X − 1 (µ1 + µ 2 ) Σ −1 (µ1 − µ 2 ) > In 2 and to ∏2 if otherwise.
2
q1
n
n
1
But q1 = 1 where n = n1 + n2 = 6 + 6 = 12 and q 2 = 2 where n = n1 + n2 = 6 + 6 = 12 , hence; q1 = q 2 =
n
n
2
1
 0.5063 − 0.5379 

DT(x) = X − 1 X 1 + X 2 S −1 X 1 − X 2 while the S −1 = 
2
 − 0.5379 0.7278 
1
0 .5
=> In(1) > 0
DT(x) = X − 1 X 1 + X 2 S −1 X 1 − X 2 > In
2
0 .5
{[
{[
{[
X1
X2
A
6 7
7 5
]
}
(
)] (
)}
(
)] (
)}
B
9
10
8
8
8 10
9 9
X1 11 15
X2 13 16
Table 4.2.1: Confusion Matrix
Statistician Decision
∏1
6
1
Population ∏1
Population ∏2
22 17 12
20 16 11
13
14
∏2
0
5
P(2/1) = 0/6 = 0
P (1/2) = 1/6
2 1
n  + n 
1
1 2
Error rate =
=
4.3: Using Bayes Criterion
The classification rule:
f1 ( x) q 2
>
and to ∏2 if otherwise.
f 2 ( x) q1
1
f1 ( x)
q
= exp X − 1 (µ1 + µ 2 ) Σ −1 (µ1 − µ 2 ) > 2
2
f 2 ( x)
q1
q
1
Note that q1 = q 2 = which means that 2 = 1
2
q1
1
q
= exp X − 1 (µ1 + µ 2 ) Σ −1 (µ1 − µ 2 ) > 2
2
q1
 0.5063 − 0.5379 

S −1 = ∑ −1 = 
 − 0.5379 0.7278 
Observation X is assigned to ∏1 if
{[
{[
}
]
]
}
{[
Therefore, observation X is assigned to ∏1 if exp X − 1
(X
2
)] (
1
1
)}
+ X 2 S −1 X 1 − X 2 > 1 and to ∏2 if otherwise.
24
Population ∏1
Population ∏2
Table 4.3.1 Confusion Matrix
Statistician Decision
∏1
6
1
∏2
0
5
P(2/1) = 0/6 = 0
P (1/2) = 1/6
2 1
n  + n 
1
1 2
Error rate =
=
= 0.0833
5.0
Summary, Conclusion and
Recommendation
5.1: Summary
Discriminant Analysis and Classification is
defined by Johnson and Wichern (1992) as
multivariate techniques concerned with separating
district set of objects and with allocating new
objects to previously defined groups.
In fisher’s criterion, object X is assigned to
Y + Y2
population ∏1 if Y > 1
and to ∏2 if
2
otherwise; and in Welch’s criterion, the optimal
rule is to assign the unit X to ∏1 if
DT(x)
=
1
q
X − 1 (µ1 + µ 2 ) Σ −1 (µ1 − µ 2 ) > In 2 and to ∏2
2
q1
if otherwise while in Bayes theorem, the object X is
{[
]
}
assigned to ∏1 if
f1 ( x) q 2
>
and to ∏2 if
f 2 ( x) q1
otherwise.
5.2: Conclusion
In other to know the best criteria among fisher’s
criterion, Welch’s criterion, and Bayes criterion, we
carried out a linear discriminant analysis on two
groups (or regions) of contrived object (or
observations). After the analysis, we discovered
that the three criteria (Fisher’s criterion, Welch’s
criterion, and Bayes criterion) had equal error rate,
that is, none of them is better than each other in
linear discriminant analysis.
5.3: Recommendation
We recommend for further studies with enlarged
sample size to ascertain if the conclusion can be
validated.
_________________________________________________________________________________
25
References
[1]Anderson T.W (1973) “Asymptotic evaluation of the probabilities of misclassification by linear discriminant
functions” In Discriminant analysis and applications, T.Cacoullos edition New York Academic
press page 17-35.
[2]Bartlelt M.S. (1951) “An inverse matrix adjustment arising in discriminant analysis” Annals of Mathematical
Statistics, 22 page 107-111.
[3]Costanza W.J. and Afifi A.A. (1979)”Comparison of stopping rules in forward stepwise discriminant
analysis” Journal of American Statistical Association, 74, page 777-785.
[4]Lachenbruch P.A. (1968) “On the expected values of probabilities of misclassification in discriminant
analysis, necessary size and a relation with the multiple correlation coefficient” Biometrics 24, page
823.
[5]Lachenbruch P.A. (1975) Discriminant Analysis. Hafner press New York.
[6] Lachenbruch P.A. and Mickey M.R. (1968) “Estimation of Error Rates in Discriminant Analysis”
Technometrics, 10, page 1.
[7] Onyeagu S.I. and Adeboye O.S. (1996) “Some methods of Estimating the probability of misclassification in
Discriminant Analysis” Journal of the mathematical Association of Nigeria ABACUS vol 24 no 2
page 104-112.
[8] Onyeagu Sidney I. (2003): “A first Course in Multivariate Statistical Analysis”, A Profession in Nigeria
Statistical Association, Mega concept Publishers, Awka, Page. 208-221.
[9] Smith C.A.B. (1947) “Some examples of discrimination” Annals of Eugenics, 18 page 272-283.
26
A Computational Analysis of the Negative Impact of Cigarette Smoking on Human
Population In Imo State
Ekwonwune E+, Osuagwu O.E.++, Edebatu D*
+Department of Computer Science, Imo State University, Owerri. Nigeria
++Department of Information Mgt Technology, Federal University of Technology, Owerri, Nigeria
*Department of Computer Science, Nnamdi Azikiwe University, Awka, Anambra State, Nigeria
Abstract
Smoking is a practice in which a substance most commonly called Tobacco or Cannabis is burnt and the smoke
tasted or inhaled. Recognition of the consequences of cigarette smoking and abuse on physical and mental
health as well as socio-occupational life are necessary steps for initiating appropriate action to reduce the
harm or dangers resulting from smoking. This work was motivated by the observed and anticipated negative
health burden with its concomitant socio-economic consequences which the nation is bound to face if
systematic efforts are not made now to control the growing problem of cigarette smoking. Three methodologies
have been combined in the execution of this research. The first methodology involved conducting the clinical
test to determine the independent assessment of impact of smoking using Digital Display Nicotine Tester
(DDNT). Secondly, sample populations of people treated at the Imo State University Teaching Hospital from
diseases emanating from smoking were collected, statistically analyzed using Statistical Packages for Social
Sciences (SPSS).Relevant coefficients were extracted and deployed for the coding of the simulation model.
Thirdly, simulation software was developed using the indices collected from the statistical software to assess
the impact of smoking on t population in the next 50 years. This is to assist policy formlators and decision
makers on what public policy should be in place to stem possible health catastrophe that may occur as a result
of uncontrolled consumption. The software simulation follows a stochastic model.
________________________________________________________________________
Introduction
The issue of smoking and associated health
risks in human beings have become a crucial
matter for discussion. Most people today
engage in one type of Tobacco smoking or the
other without knowing its negative impact on
human beings. The inhalation of products of
tobacco may cause serious respiratory
complications.
According to WHO [12] as many as onethird of patients admitted to burn treatment unit
have pulmonary injury from smoke inhalation.
Morbidity and deaths due to smoke inhalation
exceed those attributed to the burns themselves.
This same report also shows that the death rate
of patients with both severe body burns and
smoke inhalation exceeds 50%.
In 1612, six years after the settlement of
Jamestown, John Rolfe was credited as the first
settler to successfully raise tobacco as a cash
crop. The demand quickly grew as tobacco,
referred to as "golden weed", reviving the
Virginia join stock company from its failed
gold expeditions [7]. In order to meet .demands
from the old world, tobacco was grown in
succession, quickly depleting the land. This
became a motivator to settle west into the
unknown continent, and likewise an expansion
of tobacco production [3]. Indentured servitude
became the primary labor force up until Bacon's
Rebellion, from which the focus turned to
slavery. This trend abated following the
American Revolution as slavery became
regarded as unprofitable. However the practice
was revived in 1794 with the invention of the
cotton gin [7].
A Frenchman named Jean Nicot (from whose
name the word Nicotine was derived)
introduced tobacco to France in 1560. From
France tobacco spread to England. The first
report of a smoking Englishman was a sailor in
Bristol in 1556, seen "emitting smoke from his
27
nostrils” [11]. Like tea, coffee and opium,
tobacco was just one of many intoxicants that
were originally used as a form of medicine [5].
Tobacco was introduced around 1600 by
French merchants in what today is modern-day
Gambia and Senegal. At the same time
caravans from Morocco brought tobacco to the
areas around Timbuktu and the Portuguese
brought the commodity (and the plant) to
southern Africa, establishing the popularity of
tobacco throughout all of Africa by the 1650s.
Soon after its introduction to the Old World,
tobacco came under frequent criticism from
state and religious leaders. Murad IV, sultan of
the Ottoman Empire 1623-40 was among the
first to attempt a smoking ban by claiming it
was a threat to public moral and health [5]. The
Chinese emperor Chongzhen issued an edict
banning smoking two years before his death
and the overthrow of the Ming dynasty. Later,
the Manchu of the Qing dynasty, who were
originally a tribe of nomadic horse warriors,
would proclaim smoking "a more heinous crime
than that even of neglecting archery". In Edo
period Japan, some of the earliest tobacco
plantations were scorned by the shogunate as
being a threat to the military economy by
letting valuable farmland go to waste for the
use of a recreational drug instead of being used
to plant food crops [8].
The most common method of smoking today
is through cigarettes, primarily industrially
manufactured but also hand-rolled from loose
tobacco and rolling paper. Other smoking tools
include pipes, cigars, bidis, hookahs and bongs.
It has been suggested that smoking related
disease kills one half of all long term smokers
but these diseases may also be contracted by
non-smokers. A 2007 report states that about
4.9 million people worldwide each year die as a
result of smoking.[16]
Fig 1: An elaborately decorated pipe.
Source: Proctor R. N., (2000).
Smoking is one of the most common forms
of recreational drug use. Tobacco smoking is
today by far the most popular form of smoking
and is practiced by over one billion people in
the majority of all human societies. Less
common drugs for smoking include cannabis
and opium [7].
The history of smoking can be dated to as
early as 5000 BC, and has been recorded in
many different cultures across the world. Early
smoking evolved in association with religious
ceremonies; as offerings to deities, in cleansing
rituals or to allow shamans and priests to alter
their minds for purposes of divination or
spiritual enlightenment. After the European
exploration and conquest of the Americans, the
practice of smoking tobacco quickly spread to
the rest of the world. In regions like India and
sub-Saharan Africa, it merged with existing
practices of smoking (mostly of cannabis) [6].
In Europe, it introduced a new type of social
activity and a form of drug intake which
previously had been unknown. Perception
surrounding smoking has varied over time and
from one place to another; holy and sinful,
sophisticated and vulgar, a panacea and deadly
health hazard. Only relatively recently, and
primarily in industrialized Western countries-,
has smoking come to be viewed in a decidedly
negative light. Today medical studies have
proven that smoking tobacco is among the
leading causes of many diseases such as lung
cancer, heart attacks, and erectile dysfunction
and can also lead to birth defects. The inherent
health hazards of smoking have caused many
countries to institute high taxes on tobacco
products and anti-smoking campaigns are
launched every year in an attempt to curb
tobacco smoking [6].
Many ancient civilizations such as the
Babylonians, Indians and Chinese burnt incense
as a part of religious rituals, as did the Israelites
and the later Catholic and Orthodox Christian
churches. Smoking in the Americas probably
had its origins in the incense-burning
ceremonies of shamans but was later adopted
for pleasure or as a social tool. The smoking of
tobacco and various other hallucinogenic drugs
was used to achieve trances and to come into
contact with the spirit world.
Substances such as Cannabis, clarified butter
(ghee), fish offal, dried snake skins and various
pastes molded around incense slicks dates back
at least 2000 years. Fumigation (dhupa) and fire
offerings (homa) are prescribed in the
Ayurveda for medical purposes and have been
practiced for at least 3,000 years while
28
smoking, dhumrapana (literally "drinking
smoke"), has been practiced for at least 2,000
years. Before modern times, these substances
have been consumed through pipes, with stems
of various lengths or chillums [3].
Cannabis smoking was common in the
Middle East before the arrival of tobacco, and
was early on a common social activity that
centered on the type of water pipe called a
hookah. Smoking, especially after the
introduction of tobacco, was an essential
component of Muslim society and culture and
became integrated with important traditions
such as weddings, funerals and was expressed
in architecture, clothing, literature and poetry
[5].
Cannabis smoking was introduced to SubSaharan Africa through Ethiopia and the east
African coast by either Indian or Arab traders in
the 13th century or earlier or spread on the
same trade routes as those that carried coffee,
which originated in the highlands of Ethiopia
[13]. It was smoked in calabash water pipes
with terra cotta smoking bowls, apparently an
Ethiopian invention which was later conveyed
to eastern, southern and central Africa.
At the time of the arrivals of Reports from
the first European explorers and conquistadors
to reach the Americas tell of rituals where
native priests smoked themselves into such high
degrees of intoxication that it is unlikely that
the rituals were limited to just tobacco [17].
Religious leaders have often been prominent
among those who considered smoking immoral or
outright blasphemous. In 1634 the Patriarch of
Moscow forbade the sale of tobacco and
sentenced men and women who flaunted the ban
to have their nostrils slit and their backs whipped
until skin came off their backs. The Western
church leader Urban VII likewise condemned
smoking in a papal bull of 1590. Despite many
concerted efforts, restrictions and bans were
almost universally ignored. When James I of
England, a staunch anti-smoker and the author of
A Counterblast to Tobacco, tried to curb the new
trend by enforcing a whopping 4000% tax
increase on tobacco in 1604, it proved a failure,
as London had some 7.000 tobacco sellers by the
early 17th century. Later, scrupulous rulers
would realize the futility of smoking bans and
instead turned tobacco trade and cultivation into
lucrative government monopolies [10].
Fig.2 A graph that shows the efficiency of
smoking as a way to absorb any different one.
Source: Proctor R. N., (2000)
29
TABLE 1.1:
Cardiovascular Disease
Coronary artery disease
Peripheral vascular disease
Aortic aneurysm
Stroke (at younger ages)
Cancer
Lung
Larynx, oral cavity, esophagus
Bladder, kidney
Pancreas, stomach
Lung Disorders
Cancer (as noted above)
Chronic bronchitis with airflow obstruction
Emphysema
Complications of Pregnancy
Infants—small for gestations age, higher perinatal mortality
Maternal complications— placenta previa, abruptio placenta
Gastrointestinal Complications
Peptic ulcer
Esophageal reflux
Increased Risks for Cigarette Smokers
Source: Burns, (1985.)
Methodology
Three types of methods are adopted in this
work:
The first methodology involved conducting
the clinical test to determine the independent
assessment of impact of smoking. This was
carried out with Digital Display Nicotine Tester
(DDNT).
Secondly, a sample population of people treated
at the Imo State University Teaching Hospital from
diseases emanating from smoking were collected,
statistically analyzed and relevant coefficient were
deployed for the coding of the simulation model.
The third methodology which is the development of
a simulation model to predict the negative impact of
Cigarette Smoking will be discussed in a follow-up
paper to be published in the next edition of this
journal.
Steps Involved In Using Digital Display
Nicotine Tester
To use the Digital Display Nicotine test the
following steps are involved:
Step1: Warm Up Step
1:
Press the power button; the tester
is turned on with one beep sound.
2:
The tester starts the 'countdown
timer from 20 to 00 on the display
and finally the LCD displays
0.00. At this point it is ready for
Blood Nicotine Level (BNL) test.
Step 2: Test
1:
Near and blow into the breath
inhaler for seconds
2:
Read the test result on the LCD
(the level of Nicotine in the
consumer's blood in mg/L)
30
3:
•
•
•
•
The buzzer continuously sounds
alarm quickly if the Nicotine
concentration or level is up to the
level:
Over and equal to 0.05% BNC
Over and equal to 0.08% BNC
Over and equal to 0,25mg/L
Over and equal to 0.50mg/L
Step 3: Power Off
1:
Press the power button to turn off
the tester
2:
The "OFF" is displayed on the
LSD if the device is idle for 100
seconds. At that point the power
button is pressed to turn off the
equipment.
Recording Format For Experimental Result
The table of comparison for Nicotine in the
blood between the mg/l and BNC is shown
below;
Table: 2 Comparison of Experimental Result With Actual BNC Values
The level of Nicotine display in mg/l
mg/L
The level of Nicotine display in BNC
0.05
0.01
0.10
1.02
0.20
0.04
0.25
0.05S
0.30
0.06
0.40
0.08
0.55
0.11
0.60
0.12
0.70
0.14
0.80
0.16
0.90
0.18
1.00
0.20
Source: IMSUTH, ORLU
The blood Nicotine level as recorded on the LCD
display is in mg/L. The actual Blood Nicotine
Concentration (BNC) is obtained by dividing the
value on the LCD display by five (5).
The second methodology is guided by the
Structured System
Analysis
and
Design
Methodology (SSADM).
This enables the
researcher to study the existing methodology of
Nicotine testing for purposes of identifying gaps
and weaknesses in order to deploy the improved
methodology using the DDNT Clinical testing tool.
The SSADM study enables the acquisition of data
that will enhance the development of a computer
simulated solution of cigarette smoking addiction
model.
To change the current system used in
determining the Blood Nicotine Concentration
(BNC) using blood serum test, to a better method
using a device called Digital Display Nicotine
Tester (Fig. 2). It is pertinent to carry out an indepth system's investigation and analysis of the old
system. A High Level Model was developed from
the study of the present system to solve the
problems identified at the analysis stage.
The Structured Systems Analysis And Design
Methodology (SSADN) Steps
The methodology adopted here in the second
phase of the study is the Structured Systems
Analysis and Design Methodology (SSADM). The
SSADM is the standard information system
development method. It takes a top-down approach
to system development, in which a high level
picture of system requirements is built and then
31
the standard deviation and the correlation between
the various consumption of Tobacco impact
Data Source
The source of the data for the work is Imo State
University Teaching Hospital, IMUSTH, Orlu. Data
was collected on the following variable over a
period of 24 months, from the records of patients
suffering from cigarette smoking-related killer
diseases.
•
Liver disease
•
Lung disease
•
Hepatitis
•
Brain damage
From the records, the total number of deaths
resulting from these diseases was also recorded.
Under the period, about 363 patients were found to
be suffering from these diseases. Information was
also collected on age of these patients
gradually refined into a detailed and rigorous
system design. Its steps include:
i. Problem identification
ii. Feasibility study
iii. Analysis
iv. Design
v. Implementation
vi. Post implementation maintenance
Discussion of Findings
This section deals with the analysis of the
collected (secondary) data from the Imo State
University Teaching Hospital, IMUSTH, Orlu.
Regression model was developed to estimate the
death rates resulting from the tobacco smoking in
the society. The descriptive statistics of each of the
variables was calculated. These include: the mean,
Data Arrangement
The Data Collected Are Rearranged In Table 3
Table 3: Number Of Deaths From Four Killer Diseases
S/no
Month
1
Lung disease x2
Hepatitis x3
Brain damage
x4
0
0
0
19
0
1
0
18
No of
death
Liver
x1
Total No of
patients
Jan.
1
1
2
Feb.
2
1
3
March 2007
5
4
0
1
0
18
4
April
4
2
2
0
0
23
5
May
1
1
0
0
0
22
6
June
0
0
0
0
0
11
7
July
1
1
0
0
0
19
8
August
2
1
0
1
0
18
9
Sep.
4
1
0
2
1
16
10
Oct.
3
2
0
1
0
20
11
Nov.
3
0
3
0
0
18
12
Dec.
2
0
2
0
0
19
13
Jan. 2008
0
0
0
0
0
24
14
Feb
4
4
0
0
0
10
15
March
10
0
4
0
6
17
16
April
1
0
1
0
0
16
17
May
1
1
0
0
0
19
18
June
3
1
0
1
1
19
19
July
5
4
1
0
0
17
20
August
8
6
1
1
0
28
21
Sept.
5
5
0
0
0
12
22
Oct.
3
2
1
0
0
7
23
Nov
2
1
1
0
0
6
24
Dec.
1
0
1
0
0
5
Source: IMSUTH Orlu
32
Descriptive Statistics
A preliminary descriptive statistics were done on
the data collected for the period of 24 months. The
mean number of deaths, their correlation as well as
standard deviations were identified. The four
identified diseases caused by cigarette smoking
were analyzed separately. Below are the
computational formula used:
Y = aβx1 4Βx2 +… βxn
n
2
Mean (X) =
X-number of deaths in each month
Standard deviation (SD) =
(1.1)
…
And
(1.2)
n
n
n
n
Correlation (r) =
(1.3)
Where
X = no of deaths from one disease
Y
= no of deaths from another disease
Using the software “statistical packages for scientific studies
(SPSS)”, the following results were obtained.
Mean
Liver = x1
Lung disease =x2
Hepatitis =x3
Brain Damage =x4
Total number of patients x
=
=
=
=
=
Standard Deviation
1.6818
SD1
0.8636
SD2
0.3182
SD3
0.0909
SD4
16.5455
SD
Regression Analysis
A regression model proposed is a non-intercept
multiple regression model. This model will help to
explain the proportion of deaths that can be
=
=
=
=
=
1.78316
1.20694
0.56790
0.29424
5.90179
attributed to liver disease, lung disease, hepatitis
and Brain damage, out of the total number of
patients. The proposed non-intercept model is of the
form:
Y = β1 X1 + β2X2+β3X3 +β4X4 + e
Where
Y
X1
X2
X3
=
=
=
=
(1.4)
Total number of patients
number of expected deaths from liver disease
number of expected deaths from lung disease
number of expected death from hepatitis
33
X4
E
=
=
number of expected death from Brain
error term.
β1 β2 β3 and β4 are the model parameters that will be estimated using the following formulas:
Y = βX+e, y=β1x1 +β2x2+ β3x3+β4x4+e
(1.5)
Putting this in matrix form we have
Where
Y=Xβ+e
y1
X11 X12 X13 X14
Y=
y2
:
,
.
X
=
X21 X22 X23 X24
X31 X32 X33 X34
……………………….
yn
Xn1 Xn2 Xn3 … Xn4
ei
e = e2
e2
:
.
en
(1.6)
Using method of least squares, we can derive the estimate of β1 as follows:
e = Y –βX
(1.7)
It is intended to minimize the sum of squares of the error term.
SSe – e1e
=
=
(Y - X β)1 (Y-X β)
Y1Y-Y1Xβ -β1X1Y -β1X1Xβ
SSe = y1y-2β1x1y+β1x1xβ
(1.8)
Differentiating (4.6) wrt β and equating to zero, we obtain
dSSe = 2x 1y-2β1x1x=0
dβ
2β1= 2 X 1Y
β= X1Y
= (X1X)-1 (X1Y)
X1X
The fitted model will be tested for its adequacy using an ANOVA table.
(1.9)
34
.
The hypothesis of interest is:
H0: B1= B2= B3= B4=0
Ha: B1≠ B2≠ B3≠ B4≠0
Table 4:
Source of
variation
Regression
Error
Total
Where
K
=
N
=
Anova Table to Test Model Adequacy
Df
Sum of square Mean sum of
F-ratio
square
K
SSR
MSR
MSR/MSe
N-K-1
SSe
MSe
N-1
SST
number of parameters in the model
number of observations per variable
Λ
SSR
=
β1X1Y-NY2.
SSe
SST
MSR
MSe
=
=
=
=
Y1Y-βX1Y
Y1Y-NY2
SSR/K
SSe/ (N-K-1)
(1.10a)
Λ
(1.10b)
(1.10c)
(1.10d)
(1.10e)
The model is significant if F-ratio > F table at NK-1, N-1 degrees of freedom and 5% level of
significance or if the P-value given by the computer
is less than 5%.,
If the model is significant it does not mean that
all the parameters are significantly different from
zero. So, we still need to test for each individual
parameter significance using t-test given by
(X1X)
δ2
C
1 ii
=
=
MSe
Diagonal element of
The whole analyses were done using SPSS.
The results are hereby presented.
t
= βi
Se (β)
Where
= βi
(4.9)
δ2Cii
βi
=
Coefficient of intercept
Y
3.095X1+4.782X2+8.329X3+1.911X4
(4.10)
=
This model was tested for significance using
the Anova table below.
.
Table 5:
Model
Anova table presented from SPSS.
Sum of square
Df
Regression
4558.205
4
Error
Total
2195.795
6754.000
18.
22
Using a significance level of 5%, the computed
model is significant. That is, we reject Ho and
accept Ha saying that at least one of the parameters
.
Coefficient
B1 =3.095
B2 =4.782
B3 =8.329
B4 =1.911
Liver Vs Lung disease
Mean sum of
squares
1139.551
F
9.341
P-value
0.000
121.989
is significantly different from zero. To know which
of these Bi is significantly different from zero, we
have
the
following
t-test
results:
Table 6: T- Test table for test of significance
t-value
P-value
2.556
0.020
2.856
0.010
1.337
0.198
0.163
0.872
CORRELATION
Table 7: Correlation Table
R2
P-value
0.257
0.124
Remark
Significant
Significant
Not significant
Not significant
Remark
Not significant
Liver Vs. Hepatitis
0.470
0.014
Significant
Liver Vs. Brain damage
0.125
0.290
Not significant
Lung disease Vs Hepatitis
0.049
0.415
Not significant
Hepatitis Vs Brain Damage
0.000
0.5
Not significant
Hepatitis Vs Brain Damage
0.707
0.000
Significant
Discussion of Results
The above result shows that during the period
under study, an average of 1.6818 person die
monthly as a result of tobacco smoking-related liver
disease, while an average of 0.8636 person die
monthly from lung disease with standard deviations
of 1.78316 and 1.20694 respectively. The average
numbers of persons dying from hepatitis and Brain
Damage monthly with their standard deviations are
0.3182 with SD of 0.5680 and 0.0909 with SD of
0.294 respectively. This shows that more deaths are
recorded from liver disease than any other disease.
This may be due to the fact that tobacco smoking
.
directly has an impact on the liver since it makes
the liver to over work.
On the total number of patients suffering from
the four killer diseases monthly, an average of
16.5455 with standard deviation of 5.9079, this
figure is alarming since it may lead to more deaths
being recorded monthly if not checked.
Correlation shows the degree of linear (one on
one) relationship between two variables. Thus,
when a correlation value is recorded, it simply
shows the strength of linear relationship between a
pair [14 ].
There may be other strong or more powerful
relationship between the pair that is not linear. Our
analysis is purely on linear relationship. From the
above correlation table, a correlation value of 0.257
was computed between the number of deaths
recorded from liver disease and lung disease. This
value was not significant at 5% level of
significance, but shows a positive weak correlation
between the pair. This value further shows that the
number of deaths recorded from both diseases
either increases or decreases together over the
period under study. A very significant value of
correlation was recorded between liver disease and
Hepatitis. The figure of 0.470 with a P-value of
0.014 which shows that it is significant at 5%
indicates that there is a moderate positive
correlation, between the number of deaths recorded
from liver disease and that recorded from hepatitis.
Both diseases are moving in the same pattern.
Deaths recorded from liver and Brain damage had
no significant relationship as a value of 0.125 and a
P-value of 0.290 were recorded. Lung disease and
Hepatitis also recorded a poor relationship as well
as Hypertension and Brain damage which recorded
correlation values of 0.049 and 0.000 respectively.
The highly significant pair is between Hepatitis and
Brain damage which recorded a correlation value of
0.707 with a P-value of 0.000. This shows that there
is a strong positive correlation between Hepatitis
and Brain damage.
Coefficient Of Multiple Determination (R2)
The R2, tells us about the amount of variation in
Y (total number of patients) that is accounted for by
the number of death from liver, lung diseases,
Hepatitis, and Brain damage. Thus for our model,
the R2 computed is 0.603, which shows that about
60.3% of the variation in y can be accounted for by
X1,X2,X3 and X4.
Interpretation of Results
The model Y = 3.095X1 + 4.782X2 + 8.329X3 +
1.911X4 shows that for every unit death as a result
of liver disease, about 3.095 persons are patients
suffering from any of the four diseases. Also for
every unit death as a result of lung disease, about
4.782 patients are suffering from any of the four
diseases. Also, for every unit death in hepatitis,
8.329 patients are suffering from the four diseases.
In a similar manner, for every unit death as a result
of brain damage, about 1.911 persons will be
suffering from the four diseases as a result of a unit
of death from brain damage.
The test for the parameter significance shows
that only death from liver disease and lung disease
are significant at 5%. The number of deaths from
hepatitis and brain damage are not significant at
5%. This does not mean that there are no deaths
recorded in these diseases, but that the number of
deaths recorded as a result of these diseases is not
significant. This shows that B1 and B2 can be used
for future predictions with certainty, but the
prediction to be made with B3 and B4 may not be
accurate.
___________________________________________________________________
.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[9]
[10]
[11]
[12]
[13]
Doll R. and Hilly A., (1950). Smoking and carcinoma of the
lung. Preliminary report. British Medical Journal 2 (4682)
Doll R. and Hilly A., (1954). The mortality of doctors in
relation to their smoking habits: A preliminary report. British Medical Journal
Gay and Freud, (1988). A Life for Our Time. New York:
W:W. Norton and Company. Pp. 650-651.
Geyelin M., (1998). Forty-six state agree to accept and 206
billion tobacco settlement. Wall Street Journal.
Harris J. R., (1998). The nurture assumption: Why children
turn out the way they do. Free Press New York. High Versus Low Fidelity Simulations: Does the
Type of Format Affect Candidates Performance or Perceptions?
Hilton M., (2000). “Smoking in British popular culture,
1800-2000: perfect pleasures”. Manchester University Press, p. 229-241.
Iverson L, (2004). why do We smoke?: The physiology of
smoking. in smoke. p 318-320.
Pollard T, (2006). The pleasures and perils of smoking in
Early Modern England. In smoke, p. 38.
Rock VJ, Malarcher A., Kahende JW, Asman K., Husten
C., Caraballo R., et al (2007). Cigarette Smoking Among Adults United State, 2006. United
States
Centers
for
Diseases
Control
and
Prevention.
http//www.cdc.gov/mmwr/preview/mmwrhtml/htm.
Timon Screech, (2004). Tobacco in Edo Period Japan in
smoke.
Trugar J, (1994). The people’s Chronology.
West R. and Shiffman S., (2007). Fast fact: smoking
cessation.
World Health Organization Regional Office for the
Western Pacific (2007). WHO/WPRO-Tobacco Fact sheet. Retrieved 2009-01-01.
Yule, G.U and Kendall, M.G. (1950), “An introduction to
the Theory of Statistics”, 14th Edition (5th Impression 1968). Charles Griffin & Co. pp258-270
.
An Application of Path Sharing To Routing For Mobile Sinks In Wireless Sensor
Networks
Okafor Friday Onyema+, Fagbohunmi Griffin Siji*
+Department of Computer Science
Michael Okpara University of Agriculture, Umudike Tel:+234-803-7172374 Email: Revmachi_4@Yahoo.Co.Uk
*Department of Computer Science, Abia State Polytechnic, Aba Tel:+234-706-4808382. email: oluwasijibomi1@hotmail.com
Abstract
Power Conservation is one of the most important challenges in wireless sensor networks. In this paper, we
present a minimum-energy routing algorithm. Our main goal is to reduce power consumed and prolong the
lifespan of the network. The protocol, named CODEXT: Coordination-based Data dissemination for Sensor
Networks eXTension, addresses the sensor networks consisting of mobile sinks. CODEXT which is an
improvement over CODE protocol Coordination-based Data dissemination for sensor networks considers
energy conservation not only in communication but also in idle-to-sleep state. Better informed routing
decisions can often be made by sharing information among neighbouring nodes. To this end, we describe the
CODEXT protocol, a generic outline for Wireless Sensor Network (WSN) protocols that focuses on locally
sharing feedback with little or no overhead. This paper describes one instantiation of it, CODEXT protocol for
optimizing routing to multiple sinks through reinforcement learning. Such a routing situation arises in WSNs
with multiple, possibly mobile sinks, such as WSNs with actuators deployed in parallel to sensors. This protocol
is based on GAF protocol and grid structure to reduce energy consumed. Our simulation results show that
CODEXT gain energy efficiency and prolong the network lifetime.
Keywords: Source, Sink, Coordination-based Data dissemination protocol, WSNs
________________________________________________________________________________
1.0 Introduction
A wireless sensor network is randomly deployed
by hundreds or thousands of unattended and
constrained sensor nodes in an area of interest.
These networking sensors collaborate among
themselves to collect, process, analyze and
disseminate data. In the sensor networks, a data
source is defined as a sensor node that either
detects the stimulus or is in charge of sensing
requested information. The sources are usually
located where environment activities of interest
take place. A sink is defined as a user’s equipment
such as PDA, laptop, etc. which gathers data from
the sensor network.
Limitations of sensors in terms of memory,
energy, and computation capacities give rise to
many research issues in the wireless sensor
networks. In recent years, a bundle of data
dissemination protocols have been proposed [3].
Most of these efforts focus on energy conservation
due to the energy limitation and the difficulty of
.
recharging batteries of thousands of sensors in
hostile or remote environment. Generally, the
power consumption of sensors can be used for three
functionalities - the power consumed for the: (a)
transmission of packets (b) reception of packets and
(c) the power consumed when the network is idle.
Besides, recent studies have shown that radio
communication dominates energy consumption in
the sensor networks, rather than computation [7];
therefore, power conservation is an especially
important challenge at the communication layers.
Each sensor network possesses its own
characteristics to cater for different applications. An
example of such applications is the monitoring and
control of safety-critical military, environmental, or
domestic infrastructure systems. Depending on each
application, the sinks may be mobile while the
sensors are stationary. On the other hand, the
number of sinks may be large since many users
may simultaneously access the sensor networks. In
this paper, we propose an energy-efficient data
dissemination approaches which have been built as
an improvement over the CODE protocol. These
protocols individually address the sensor networks
consisting of mobile sinks and the sensor networks
consisting of a large number of sinks.
The algorithm, Coordination-based Data
Dissemination Protocol Extension (or CODEXT for
short), addresses mobile sinks. The authors are
motivated by the fact that handling mobile sinks is
a challenge of large-scale sensor network research.
Though many researches have been published to
provide efficient data dissemination protocols to
mobile sinks [9]; they have proposed how to
minimize energy consumed for network
communication, regardless of
idling energy
consumption. In fact, energy consumed for nodes
while idling cannot be ignored [10], show that
energy consumption for idle:receive:send ratios are
1:1.05:1.4, respectively. Consequently, they
suggest that energy optimizations must turn off the
radio. Doing this not only simply reduces number
of packets transmitted but also conserves energy
both in overhead due to data transfer, and in idle
state energy dissipation when no traffic exists,
especially in sensor networks with high node
density. In CODEXT, we take into account the
energy for both communication and idle states.
CODEXT provides an energy efficient data
dissemination path to mobile sinks for coordination
sensor networks. CODEXT is based on grid
structure and coordination protocol GAF [13]. The
key observation driving the CODEXT notion is that
wireless communication between neighbouring
nodes is not a private, point-to-point exchange
between them, but rather it is broadcast, implying
that it can be received by all nodes within range.
Extensive amounts of local data exist on the single
nodes in a wireless network, which, if shared, could
improve the performance of the routing and or
application levels. This data is usually small, such
as residual energy, available routes to sinks, route
costs to specific sinks, application role assigned to
the node, link quality information, etc. When shared
with neighbours, this information could be used for
adjusting existing routes and taking routing
decisions to minimize costs.
To better understand the rest of the paper, the
authors first describe the general protocol design
goals of sensor networks in Section 2. Then in
section 3 and 4, we present the protocol and its
performance evaluation.. The discussion about
benefit of the proposed approach is given right after
its evaluation. Section 5 concludes the paper.
1.1 Protocol Design Goals
The wireless sensor network has its own
constraint that differs from adhoc networks. Such
.
constraints make designing routing protocol for
sensor networks very challenging [1]. Firstly,
sensor nodes are limited in power, processing
capacities and memory. These require careful
resource management. Secondly, sensor nodes may
not have global identifications (IDs). Therefore,
classical IP-based protocol can not be applied to the
sensor networks. Thirdly, sensor nodes might be
deployed densely in the sensor networks.
Unnecessary nodes should turn off its radio while
guaranteeing connectivity of the entire sensor field.
Fourthly, generated data traffic has significant
redundancy in it since multiple sensors may
generate same data within the vicinity of a
phenomenon. Such redundancy needs to be
exploited (through compression techniques) by the
routing protocols to improve energy and bandwidth
utilization. This will be addressed in the clustering
algorithm to be proposed later.
In order to design a good protocol for the sensor
networks, such constraints should be managed in an
efficient manner. In this paper, emphases was
placed on three major design goals in data
dissemination protocol for wireless sensor
networks.
1.1.1 Energy Efficiency/Network Lifetime
Energy efficiency is the most important
consideration due to the power constraint of sensor
nodes. Recent studies have shown that radio
communication is the dominant consumer of energy
in the sensor networks. Most of recent publications
mainly focus on how to minimize energy
consumption for sensor networks. Besides, multihop routing will consume less energy than direct
communication, since the transmission power of a
wireless radio is proportional to the distance
squared or even higher order in the presence of
obstacle. However, multi-hop routing introduces
significant overhead for topology management and
medium access control [1]. Another characteristic
of the common sensor networks is that sensor nodes
usually generate significant redundant data.
Therefore similar packets from multiple nodes
should be aggregated so that the number of packets
transmitted would be reduced [8]. Several work,
[7], [11], suggest that unnecessary nodes should be
turned off to conserve energy and reduce collision.
1.1.2 Latency
The user is interested in knowing about the
phenomena within a given delay. Therefore, it is
important to receive the data in a timely manner [5],
[7].
1.1.3 Scalability
Scalability is also critical factor. For a large scale
sensor network, it is likely that localizing
interactions through hierarchical and aggregation
will be critical to ensure scalability [5]. Keeping
these design goals in mind, in this paper we propose
a data dissemination protocols for large-scale
sensor networks to achieve energy efficiency while
guaranteeing a comparable latency with existing
approaches.
1.2 CODEXT: A Coordination-Based Data
Dissemination Protocols To Mobile Sink
CODEXT addresses the sensor networks
consisting of mobile sinks. In CODEXT, we rely on
the assumptions that all sensor nodes are stationary.
Each sensor is aware of its residual energy and
geographical location. Once a stimulus appears, the
sensors surrounding it collectively process the
signal and one of them becomes the source to
generate data report. The sink and the source are
not supposed to know any a-priori knowledge of
potential position of each other. To make
.
unnecessary nodes stay in the sleeping mode,
CODEXT is deployed above GAF-basic protocol
[10]. Fig.1 depicts CODE general model where the
routing algorithm is implemented above the GAF
protocol. In this paper, we only focus on CODEXT
routing algorithm. Details of GAF algorithm can be
referred in [13].
Fig.1.CODEXT System Model
The basic idea of CODEXT is to divide sensor
field into grids. Grids are indexed based on its
geographical location. According to GAF, each grid
contains one coordinator which acts as an
intermediate node to cache and relay data.
CODEXT consists of three phases: data
announcement,
query
transfer
and
data
dissemination. As a stimulus is detected, a source
generates a data-announcement message and sends
to all coordinators using simply flooding
mechanism. Each coordinator is supposed to
maintain a piece of information of the source
including the stimulus and the source’s location. As
a mobile sink joins in the network, it selects a
coordinator in the same grid to act as its Agent.
When it needs data, it sends a query to this Agent.
The Agent is in charge of forwarding the query to
the source based on the target’s location and grid
IDs. An efficient data dissemination path is
established while the query traverses to the source.
Receiving a query, the source sends the data to the
sink along the data dissemination path. The Agent
helps the sink to continuously keep receiving data
from the source when the sink moves around.
Periodically, the sink checks its location. If the sink
moves to another grid, it first sends cache-removal
message to clear out the previous data
dissemination path and then re-sends a query to
establish a new route.
1.3 CODEXT Theory
A. Grid Indexing
We assume that we have partitioned the network
plane in virtual MxN grids (for example in Fig.2
that is 3x2 grids). Each grid ID which has a typed
[CX.CY] is assigned as follows: at the first row,
from left to right, the grid IDs are [0.0], [1.0], and
[2.0]. Likewise, at the second row, grid IDs are
[0.1], [1.1], and [2.1] and so forth. To do this, based
on the coordinate (x, y), each node computed itself
CX and CY:
(1)
.
where r is the grid size and [x] is largest integer less
than x.
Fig.2.Grid Indexing
B. CODEXT Algorithms
a) Data Announcement
When a stimulus is detected, the source
propagates a data-announcement message to all
coordinators using simply flooding mechanism.
Every coordinator stores a few piece of information
for the data dissemination path discovery, including
the information of the stimulus and the source
location. In this approach, the source location does
not mean the precise location of the source, but its
grid ID. Since the coordinator role might be
changed every time, the grid ID is the best solution
for nodes to know the target it should relay the
query to. To avoid keeping data-announcement
message at each coordinator indefinitely, a source
includes a timeout parameter in data-announcement
message. If this timeout expires and a coordinator
does not receive any further data-announcement
message, it clears the information of the stimulus
and the target’s location to release the cache.
b) Query Transfer
Every node is supposed to maintain a Query
INformation Table (hereafter called QINT) in its
cache. Each entry is identified by a tuple of (query,
sink, uplink) (sink is the node which originally
sends the query; uplink is the last hop from which
the node receives the query). By definition, two
entries in QINT are identical if all their
corresponding elements are identical. For example
in Fig.3, node n1 and node n2 receive a query from
sink1 and sink2, therefore it maintains a QINT as
Fig.4.
Fig.3.Query Transfer And Data Dissemination
Path Setup
Fig.4.Query Information Table Maintained At Nodes n1 and n2
.
Receiving a query from an uplink node, a node
first checks if the query exists in its QINT. If so, the
node simply discards the query. Otherwise, it
caches the query in the QINT. Then, based on
target’s location stored in each coordinator, it
computes the ID of next grid to forward the query.
This algorithm is described in Fig.5. In this figure,
NODE is the current node handling the query
packet and src_addr contains the target’s location.
If NODE is the source, it starts sending data along
the data dissemination path. Otherwise, it finds the
next grid which is closest to the target to relay the
query. In case the next grid contains no node (socalled void grid) or the next grid’s coordinator is
unreachable, it tries to find a round path. To do this,
it first calculates the disparity, δCX, δCY_..
The next grid will be
NextGrid.CX = NODE.CX δCX
NextGrid.CY = NODE.CY δCY
Fig.5.Pseudo-Code Of Finding Next Grid ID Algorithm
Each node is supposed to maintain a one-hopneighbour table. (i.e. information about its one-hop
neighbours). If a node can not find the next grid’s
.
coordinator in this table, it considers that grid as a
void
grid
.
Fig.6.Multi-Hop Routing Through Coordinators
For example in Fig.6, the sink1 sends query to the
source src along the path [4.1], [3.2], [2.3], [1.3],
[0.3]. However, with the sink2, the grid [3.0]’s
coordinator can not find grid [2.1]’s neighbour (due
to void grid) and grid [3.1]’s coordinator also can
not find grid [2.2]’s neighbor (due to unreachable
node) in its one-hop-neighbour table. Therefore, it
finds the round path as [3.1], [3.2], [2.3], [1.3],
[0.3]. A data dissemination path is discovered by
maintaining a QINT at each intermediate node. A
query from a sink is re-transmitted when the sink
moves to another grid. The path length of each
neighbour for each sink are stored in a Neighbour
Table, e.g.,
1: init:
2: CODEXT.init();
3: routeData(DATA):
4: CODEXT.updateFitness(DATA.Routing,
.
c) Data Dissemination
A source starts generating and transmits data to a
sink as it receives a query. Upon receiving data
from another node, a node on the dissemination
path (including the source) first checks its QINT if
the data matches to any query to which uplinks it
has to forward.
If it finds that the data
5:
6:
7:
8:
9:
10:
11:
12:
13:
DATA.Feedback);
if (myAddr in Routing)
if (explore)
possRoutes =
PST.getAllRoutes(DATA.Routing.sinks);
route = explore.select(possRoutes);
else
route =
CODEXT.getBestRoute(DATA.Routing.sin
ks);
DATA.Feedback.value =
CODEXT.getBestCost(DATA.Routing.sink
s);
DATA.Routing = route;
sendBroadcast(DATA);
Fig. 7. CODEXT Pseudo Code Initialization And
Processing Of One DATA Packet
matches several queries but with the same uplink
node, it forwards only one copy of data. Doing this
reduces considerable amount of data transmitted
throughout the sensor network. For example in
Fig.4, node n1 receives the same query A of sink1
and sink2 from the same uplink node (n2).
.
Therefore, when n1 receives data, it sends only one
copy of data to n2. Node n2 also receives the same
query A of sink 1 and sink 2 but from different
uplink nodes (n3, n4). Thus, it must send two
copies of data to n3 and n4. Likewise, the data is
relayed finally to the sinks.
2.0 Handling Sink Mobility
CODEXT is designed for mobile sinks. In this
section, the authors describe how a sink keeps
continuously receiving updated data from a source
while it moves around within the sensor field.
Periodically, a sink checks its current location to
know which grid it is located. The grid ID is
computed by the formula (1). If it is still in the
same grid of the last check, the sink does nothing.
Otherwise, it first sends a cache-removal message
to its old Agent. The cache-removal message
contains the query’s information, the sink’s
identification and the target’s location. The old
Agent is in charge of forwarding the message along
the old dissemination path as depicted in Fig.8.
After receiving a cache-removal message, a node
checks its QINT and removes the matched query.
When this message reaches the source, the whole
dissemination path is cleared out, i.e. each
intermediate node on the path no longer maintains
that query in its cache. Consequently, the source
stops sending data to the sink along this
dissemination path. After the old dissemination
path is removed, the sink re-sends a query to the
target location. A new dissemination path is
established as described in section (b) above. By
doing this, the number of queries which is needed
to be re-sent is reduced significantly compared with
other approaches. Hence, collision and energy
consumption is reduced. Also, the number of lost
data packet is decreased. In case the sink moves
into a void grid, it selects the closest coordinator to
act as its Agent.
Fig.8.Handling Sink Mobility
2.1 CODEXT Performance
A. Simulation Model
Here, the authors developed a simulator based on
OMNET++ simulator to evaluate and compare
CODEXT to other approaches such as Directed
Diffusion (DD) and CODE. To facilitate
comparisons with CODE and DD, we use the same
energy model used in n2 that requires about 0.66W,
.
0.359W and 0.035W for transmitting, receiving and
idling respectively. The simulation uses MAC IEEE
802.11 DCF that OMNET++ implements. The
nominal transmission range of each node is 250m,
[13].
Our goal in simulating CODEXT is to examine
how well it actually conserves power, especially in
dense sensor networks. In the simulation, we take
into account the total energy consumed for not only
transmitting, receiving but also idling. The sensor
network consists of 400 sensor nodes, which are
randomly deployed in a 2000mx2000m field (i.e.
one sensor node per 100mx100m grid). Two-ray
ground is used as the radio propagation model and
an omni-directional antenna having unity gain in
the simulation. Each data packet has 64 bytes,
query packet and the others are 36 bytes long. The
default number of sinks is 8 moving with speed 10
m/sec (i.e. the fastest human speed) according to
random way-point model (David B, J and David
A.M 1996). Two sources generate different packets
at an average interval of 1 second. Initially, the
sources send a data-announcement to all
coordinators using flooding method. When a sink
needs data, it sends a query to its Agent. As a
source receives a query, it starts generating and
sends data to the sink along the data dissemination
path. The simulation lasts for 200 seconds. Four
metrics are used to evaluate the performance of
CODEXT. The energy consumption is defined as
the total energy network consumed. The success
rate is the ratio of the number of successfully
received packets at a sink to the total number of
packets generated by a source, averaged over all
source-sink pairs. The delay is defined as the
average time between the time a source transmits a
packet and the time a sink receives the packet, also
averaged over all source-sink pairs. We define the
network lifetime as the number of nodes alive over
time.
2.2 Performance Results
a) Impact of Sink Number
The impact of the sink number on CODEXT is
first of all studied. In the default simulation, we set
the number of sink varying from 1 to 8 with the
max speed 10m/s and a 5-second pause time.
CODEX
CODE
T
Fig.8.Energy Consumption For Different
Numbers Of Sinks
Fig.8 shows total energy consumption of
CODEXT. It demonstrates that CODEXT is more
energy efficient than other source protocols. This is
because of three reasons. Firstly, CODEXT uses
QINT to efficiently aggregate query and data along
data dissemination path. This path is established
based on grid structure. Hence CODEXT can find a
nearly straight route between a source and a sink.
Secondly, CODEXT exploits GAF protocol, so that
nodes in each grid negotiate among themselves to
turn off its radio. Thirdly CODEXT uses the
concept of SHARING TREE. The goal in
CODEXT is to route the data to multiple sinks.
Because standard routing tables show single sink
routes, we need a new data structure to manage
options for routing data through different
.
combinations of neighbours to reach different
combinations of multiple sinks. For this, we use the
CODEXT Sharing Tree, a data structure that allows
for easy selection of the next hop(s) for a given set
of sinks. The name CODEXT sharing tree derives
from the tree shape of the data structure, as well as
our goal to allow a single packet to share a path as
it travels to multiple sinks. This section outlines the
key properties of the CODEXT [5].
Figure 9: The Neighbour Table For A Sample “Home
Node” And Part Of Its Corresponding CODEXT Sharing
Tree
2.2.1 Functionality Of The CODEXT Sharing
Tree
The CODEXT sharing tree is maintained at each
node to keep all possible routes to all sinks through
all combinations of neighbours. It is worth noting
that each node, referred to as the home node in its
CODEXT sharing tree, maintains only one sharing
tree, independent of the number of sources, sinks,
and neighbours. Here we explore the CODEXT
sharing tree through its interface. init(): The
CODEXT sharing tree is initialized with data
contained in the Neighbour Table. Here, we
illustrate the CODEXT sharing tree contents
through the small example in Figure 9 where the
home node has 2 neighbours, N1 and N2, and the
system has 3 sinks, A, B, and C. The intention is to
use the CODEXT sharing tree to select the
neighbours to serve as the next hop(s) for each of
the destination sinks. As the goal is to share the
routes as much as possible, the options of using a
single neighbour to serve multiple sinks are
considered. To illustrate the routing choices
available, we observe that N1 can route packets
toward any of the following neighbour
combinations: {A}, {B}, {C}, {A,B}, {A,C},
{B,C}, {A,B,C}. The same subsets can be reached
through N2. To select the next hops for all sinks,
we must choose sets of these neighbour
combinations, such that their union includes all
desired sinks exactly once. For example, to route a
packet to all three sinks, we could select {A,B}N1
and {C}N2, where the subscript indicates the
neighbour to which the combination belongs.
Alternately, {A,B,C}N1 is sufficient. The set of all
possible routes for all three sinks is the brute force
combination of all neighbour combinations. To
structure these choices, a tree is constructed where
each node is a neighbour combination. In this tree,
a path from any leaf to the root represents a routing
option to reach all of the sinks. For example, in
Figure 9, the path from the first leaf to the home
node (the tree’s root) corresponds to the first
selection above. The final initialization step
annotates each node of the CODEXT sharing tree
with it fitness value, update Fitness(route, f). As
previously observed, fitness values are initial
estimates which are updated as the system receives
new fitness values through the feedback mechanism
of the CODEXT FRAMEWORK.
Therefore, whenever a packet is overheard, its
feedback values are used to update the
corresponding neighbour combinations, the node(s)
in the CODEXT sharing tree. update Tree(): Each
time the Neighbour Table changes due to the
insertion or deletion of a neighbour or sink, the
CODEXT sharing tree must be updated. Since the
fitness values are calculated only at initialization
.
and updated later through feedback, it is important
not to lose them during an update. Therefore, rather
than rebuild the CODEXT sharing tree from
scratch, an update function that makes only the
required changes is provided.
GetAllRoutes(sinkSet)
Every packet carries the subset of sinks that it
should be routed to by the receiving node. The
CODEXT sharing tree has the responsibility to take
this subset and enumerate the possible options for
the next hop(s). These options can be visualized as
a set of partial paths in the CODEXT sharing tree
starting at the home node. Each path must include
PST nodes, which union includes exactly the
destination sinks. getBestRoute(sinkSet): During
the stable phase of our CODEXT protocol, we
rotate among all available best routes for a specified
sink subset. For convenience, we place the
responsibility for balancing the selection among
multiple options inside the CODEXT sharing tree,
providing a single function that returns only one
route. Therefore, it reduces significantly energy
consumption. In contrast, DD (Direct Diffusion)
always propagates the new location of sinks
throughout the sensor field in order for all sensor
nodes to get the sink’s location. In CODE, the new
multi-hop path between the sink and the grid is
rebuilt. Also, data dissemination path of CODE is
along two sides of a right triangle.
Fig.10 demonstrates the average end-to-end
delay of CODEXT. As shown in this figure, the
delay of CODEXT is shorter than CODE and
slightly longer than DD. In Fig.10, it shows that the
success rate of CODEXT is always above 90
percent. It means that CODEXT delivers most of
data successfully to the multiple sinks.
CODEX
T
CODE
DD
CODEX
T
CODE
DD
Fig.10 .Delay For Different Numbers Of Sinks
Fig.11. Success Rate For Different Numbers Of
Sinks
.
2.2.2 Impact of Sink Mobility
In order to examine the impact of sink mobility,
CODEXT is measured for different sink speeds (0
to 30 m/sec). In this experiment, the network
consists of 8 mobile sinks and 400 sensor nodes.
CODEXT
CODE
Fig.13. Delay For Different Sink Speeds
rate is also above 90 percent. These results show
that CODEXT handles mobile sinks efficiently.
2.2.3 Impact Of Node Density
To evaluate the impact of node density on
CODEXT, we vary the number of nodes from 300
(1 node/cell on average) to 600 nodes (2
nodes/cell). Eight sinks move with speed 10m/sec
as default. Fig.15 shows the energy consumption at
different node densities. In this figure, CODEXT
demonstrates better energy consumption than other
protocols. As the number of nodes increase, the
total energy consumption slightly increases. This is
because of turning off node’s radio most of the
time. Therefore, energy is consumed mostly by the
coordinators. While in CODE and DD, nodes which
CODEXT
CODE
don’t participate in communication
still consume
energy in sleeping mode.
CODE
Fig.14. Success Rate For Different Sink Speeds
Fig.15 .Energy Consumption For Different Node
Density
Fig.12 demonstrates total energy consumed as
the sink speed changes. In both low and high speeds
of the sinks, CODEXT shows the total energy
consumed is better than other protocols, about twice
less than CODE and three times less than DD. The
reason is that, aside from above explanations, the
mobile sink contact with the coordinator to receive
data while it is moving. Thus, the query only needs
to resend as it moves to another grid. Fig.13 shows
the delay of CODEXT which is comparable with
CODE and longer than DD. In Fig.14, the success
2.2.4 Network Lifetime
In this experiment, the number of sinks is 8
moving with speed 10 m/sec. The number of sensor
nodes is 400. A node is considered as a dead node if
its energy is not enough to send or receive a packet.
Fig.15 shows that number of nodes alive of
CODEXT is about 60 percent higher than CODE at
the time 600sec. This is due to two reasons: The
first is that CODEXT focuses on energy efficiency.
The second is that rotating coordinators distribute
energy consumption to other nodes, thus nodes will
.
not quickly deplete its energy like other approaches.
CODEXT concentrates on dissemination nodes to
deliver data, therefore such nodes will run out of
energy quickly. We do believe that when the node
density is higher, the lifetime of CODEXT will be
prolonged much more than other approaches.
CODEXT
CODE
Fig.16. Number Of Node Alive Over Time
2.3 Conclusion
Many sensor network protocols have been
developed in recent years. [2], [4], [12]. One of the
earliest work, SPIN [3] addresses efficient
dissemination of an individual sensor’s observation
to all the sensors in the network. SPIN uses metadata negotiations to eliminate the transmission of
redundant data. Directed Diffusion [3] and DRP [7]
are similar in that they both take the data-centric
naming approach to enable in-network data
aggregation. In Directed Diffusion, all nodes are
application-aware. This enables diffusion to achieve
energy saving by selecting empirically good paths
and by caching and processing data in-network.
DRP exploits application-supplied data descriptions
to control network routing and resource allocation
in such a way as to enhance energy efficiency and
scalability. GRAB [14] targets at robust data
delivery in an extremely large sensor network made
of highly unreliable nodes. It uses a forwarding
mesh instead of a single path, where the mesh’s
width can be adjusted on the fly for each data
packet. GEAR [14], uses energy aware neighbour
selection to route a packet towards the target
region. It uses Recursive Geographic Forwarding or
Restricted Flooding algorithm to disseminate the
packet inside the destination regions.
While such previous work only addresses the
issue of delivering data to stationary sinks, other
work such as CODE [6], SEAD [2] and SAFE [9],
[3] target at efficient data dissemination to mobile
sinks. CODE exploits local flooding within a local
cell of a grid which sources build proactively. Each
source disseminates data along the nodes on the
grid line to the sink. However, it does not optimize
the path from the source to the sinks. When a
source communicated with a sink, the restriction of
grid structure may multiply the length of a straight
line path by 2. Also, CODE frequently renews the
entire path to the sinks. It therefore increases
energy consumption and the connection loss ratio.
SAFE uses flooding that is geographically limited
to forward the query to nodes along the direction of
the source. SAFE uses geographically limited
flooding to find the gate connecting itself to the
tree. Considering the large number of nodes in a
sensor networks, the network-wide flooding may
introduce considerable traffic. Another data
dissemination protocol, SEAD, considers the
distance and the packet traffic rate among nodes to
create near-optimal dissemination trees. SEAD
strikes a balance between end-to-end delay and
power consumption that favors power savings over
delay minimization. SEAD is therefore only useful
for applications with less strict delay requirements.
CODEXT differs from such protocols in three
fundamental ways. First, CODEXT exploits GAF
protocol [13] to reduce energy consumption and
.
data collision while the nodes make decision to fall
maintaining an Agent to relay data to the sink when
into sleeping mode. Second, based on grid
it moves within a grid. In addition, CODEXT takes
structure, CODEXT can control the number of
into account of query and data aggregation [5], [6]
transmitted hops and disseminates data along a path
to reduce the amount of data transmitted from
multiple sensor nodes to sinks like other
shorter than others such as CODE. Third, the
number of re-transmitted queries is reduced by
approaches.
___________________________________________________________________________
References
[1] Fan Ye et al (2002) “Sensor Networks: A Two-Tier Data Dissemination Model For LargeScale Wireless Sensor Networks” Proceedings of the Eighth Annual ACM/IEEE International Conference on
Mobile Computing and Networks (MobiCOM 2002), Atlanta, GA.
[2] Hyung Seok Kim et al (2003) “Dissemination: Minimum-Energy Asynchronous
Dissemination To Mobile Sinks In Wireless Sensor Networks” Proceedings of the first international
conference on Embedded networked sensor systems.
[3] Intanagonwiwat C et al (2003: 2-16) “Directed Diffusion For Wireless Sensor Networking”
Networking, IEEE/ACM Transactions Vol 11 Issue.1.
[4] Joanna Kulik et al (2002), “Negotiation-Based Protocols For Disseminating Information In
Wireless Sensor Networks” ACM Transaction on Vol 8 , Issue 2.
[5] Krishnamachari B, Estrin D, and Wicker S . (2002) “The Impact Of Data Aggregation In
Wireless Sensor Networks”. Proceedings of the 22nd International Conference on Distributed Computing
Systems Workshops.
[6] Maddes S et al (2002) “Supporting Aggregate Queries Over Ad-Hoc Wireless Sensor
Network”. IEEE Workshop on Mobile Computing Systems and Applcation. .
[7] Nirupama B et al (2000:28-34), “Gps-Less Low Cost Outdoor Localization For Very Small
Devices”. IEEE Personal Communications Magazine, Vol 7.
[8] Pottie G J and Kaiser W J (2000:51-58). “Embedding The Internet: Wireless Integrated
Network Sensors”. Communications of the ACM, Vol 43.
[9] Sooyeon Kim et al (2003:228-234); “A Data Dissemination Protocol For Periodic Updates In
Sensor Networks” Workshops, Proceedings. 23rd International Conference on Distributed Computing Systems.
[10] Stemm M and Katz R H. (1997) “Measuring And Reducing Energy Consumption Of
Network Interfaces In Hand-Held Devices”. IEICE Transaction and communication.
[11] Wendi B et al (1995) “An Application-Specific Protocol Architecture For Wireless
Microsensor Networks” IEEE transactions on wireless communications.
[12] Wensheng Zhang et al (2003:305-314) ”Data Dissemination With Ring-Based Index For
Wireless Sensor Networks” Proceedings. 11th IEEE International Conference on Wireless Netwoking.
[13] Xu Y et al (2001), “Geography-Informed Energy Conservation For Ad Hoc Routing”.
Proceedings . of the Seventh Annual ACM/IEEE International Conference on Mobile Computing and Networking
(MobiCom 2001), Rome, Italy.
[14] Yan Yu et al ,(2001) “Geographical And Energy Aware Routing: A Recursive Data
Dissemination Protocol For Wireless Sensor Networks”, UCLA Computer Science Department.
.
Expert System for Diagnosis of Hepatitis B
Ibrahim Mailafiya, Fatima Isiaka
Department of Computer Science,
Nasarawa State University, Keffi
Abstract
This paper is a preview of the work so far concluded on Expert Systems implementation for the diagnosis of
hepatitis B, which is one of the most common of all hepatitis ravaging mankind today. A user friendly
application programme has been developed which can diagnose and prescribe solutions to the treatment of
hepatitis B virus. The applications software has the capacity to monitor patients. The programme is limited to
diagnosis and treatment of hepatitis B virus in Nigeria and the general terms in hepatitis B are considered
under clinical study, general considerations, mechanisms regulation, virology, diagnosis and treatment.
Key Words: Expert Systems, diagnosis, treatment, clinical study, patients
_______________________________________________________________________________
interesting to the medical professionals compared to
the implementation of an already existing system.
This project will require the involvement of two
1.0 Introduction
professionals which are the medical doctors and the
The use of computers has brought tremendous
computer scientist. They can help in providing a
development to the world thereby making things a
software application package that the medical
lot easier for people to handle. In recent times, only
specialist will need using information the medical
a few parts of the world are not making good use of
specialist provides.
computers to the full. It is only in these parts of the
Thus, the computer scientist is involved in
world that may not apply this software due to high
tedious humanitarian services as well as working
cost of automation and low IT illiteracy level.
under stringent conditions such as budget and many
In the health sector two or more doctors cannot
unappreciative clients. Despite all these draw backs,
prescribe the same drug to a patient on a particular
it presents an opportunity for hepatitis patent to get
illness. Thus, the computer also at the end of the
diagnosed and treated with the help of a computer.
day will be making a routine decision for every
In a paper presented by Shikhar, he proposed an
patient. So far, there has been little success in
architectural framework of an Expert System in the
achieving transfers of technology between medical
area of agriculture and describes the design and
computing research and the real application in
development of the Rule-based Expert System,
medical science environment. In most cases new
using the Shell ESTA (Expert System for Text
researches tend to be more
Animation). The designed system is intended for
the diagnosis of common diseases occurring in the
rice plant [2].
.
An Expert System is a computer program
normally composed of a knowledge base, inference
engine and user-interface [3]. The proposed expert
system facilitates different components including
decision support module with interactive user
interfaces for diagnosis on the basis of response(s)
of the user made against the queries related to
particular disease symptoms. ESTA programming
is based on logic programming approach. The
system integrates a structured knowledge base that
contains knowledge about symptoms and remedies
of diseases in the rice plant appearing during their
life span [1].
Ali (2010 ) had designed a Fuzzy Expert
System for heart disease diagnosis. The designed
system was based on the V.A. Medical Center,
Long Beach and Cleveland Clinic Foundation data
base. The system has 13 input fields and one output
field. Input fields are chest pain type, blood
pressure, cholesterol, fasting blood sugar,
maximum heart rate, resting electrocardiography
(ECG), exercise, old peak (ST depression induced
by exercise relative to rest), thallium scan, sex and
age. The output field refers to the presence of heart
disease in the patient. It is integer valued from 0 (no
presence) to 4 (distinguish presence (values 1, 2, 3,
4)). This system uses Mamdani inference method
[4]. The results obtained from designed system are
compared with the data in the database and
observed results of designed system are correct in
94% of cases. The system was designed in Matlab
software. The system can be viewed as an
alternative for existing method.
The use of computer technology in the fields of
medicine area diagnosis, treatment of illnesses and
patient pursuit has highly increased [6].Despite the
fact that these fields, in which the computers are
used, have very high complexity and uncertainty
and the use of intelligent systems such as fuzzy
logic, artificial neural network and genetic
algorithm have been developed [5].
In the domain fields of heart disease risk, smoke,
cholesterol, blood pressure, diabetes, sex and age
are main risk factors that affect heart disease risk
[4].
Detecting diseases at early stage can enable a
patient to have early treatment which can lead to
effective control. Identifying the treatment
accurately depends on the method that is used in
diagnosing the diseases [7]. A Diagnosis expert
system (DExS) can help a great deal in identifying
those diseases and describing methods of treatment
to be carried out taking into account the user
capability in order to deal and interact with expert
system easily and clearly. Present expert system
uses inference rules and plays an important role that
will provide certain methods of diagnosis for
treatment [8].
Expert System can also be applied in Car failure
detection. It is a complicated process and requires
high level of expertise. Any attempt of developing
an expert system dealing with car failure detection
has to overcome various difficulties. The paper in
the journal describes a proposed knowledge-based
system for car failure detection [9].
A web-based expert system for wheat crop was
also developed in Pakistan. Wheat is one of the
major grain crops in Pakistan. It is cultivated in vast
areas of Punjab followed by Sindh and ranked first
as a cereal crop in the country[11]. Rule-based
expert system covers two main classes of problems
namely diseases and pests, normally encountered
in wheat crop. The expert system is intended to
help the farmers, researchers and students and
provides an efficient and goal-oriented approach
for solving common problems of wheat. The
system gives results that are correct and consistent
[10].
.
This paper is based on Hepatitis B, which is one
of the most common of all hepatitis in Nigeria.
Hepatitis B is irritation and swelling of the liver due
to infection with the hepatitis B virus - HBV.
Hepatitis B may be acute or chronic, the acute
hepatitis B last less than six months, and it may
lead to various infections that affect the liver. The
chronic hepatitis B is at the risk of a lasting liver
disease. It continues after and may persist beyond
six months.
Most of the damages from hepatitis B virus occur
because of the way the body responds to the
infection, when the body’s immune system detects
the infection it sends out special cells to fight it off,
however, these disease fighting cells can lead to
liver inflammation. Hepatitis B is also known as
Serum hepatitis.
It has been in existence for over a thousand
years. T disease has been recorded to have had a
large number of deaths in most developed
countries. The liver being the second largest organ
in the body plays an important role in regulating the
composition of various chemical cells in the body,
this is the reason why special attention should be
given to the liver. The computer which has already
stored the relevant information needed by the
physician and may be referred to as the stethoscope
that assists the medical doctors do a very good job.
Doctors can direct questions to the computer and
receive answers on what they need to judge the
diseases on the screen of the computer. This helps
the doctor draw up an effective treatment chart
thereby improving his efficiency on time, number
of patients attended to and save more lives.
1.0.1 Types of Hepatitis B
There are three types of hepatitis B
• Healthy chronic carrier of hepatitis B:
these carriers are not infectious to other people
although they may slightly have a higher risk of
cirrhosis and liver cancer. The virus becomes
reactivated when the immune system becomes
suppressed.
• Chronic infectious hepatitis B: here the
person is highly infectious to people around,
they have very inflamed and damaged liver
even when the person has few or no symptoms.
• Chronic mutant hepatitis B: here the person
has a mutant strain. A permanent alteration of
HBV genetic make. They have the potential to
be infectious to other and it is thought to be
more resistant to treatment than the other types.
1.0.2 Mode of Transmission of HBV
Hepatitis B infection can be spread through
having contact with the blood, semen, virginal
fluids and other body fluids of someone who
already has hepatitis B infection.
Infection can be spread by these modes
• Blood transfusions
• Direct contact with blood in health care
settings
• Sexual contact with an infected person.
• Tattoo and acupuncture with unclean
needles or instruments.
• Sharing needles during drug use.
• Sharing personal items such as
toothbrush, razors and nail clippers with
the infected person.
• Pains on the right side of the abdomen.
It can also be passed from mother to child during
child birth.
1.0.3 Symptoms of HBV
If the body is able to fight off the hepatitis B
virus, any symptoms should go away over a period
of weeks to 6 months. Many people with chronic
hepatitis have few or no symptoms. They may not
.
even look sick; as a result they may not know they
are infectious. Some symptoms may not appear for
up to 6 months after the infection. Early symptoms
may include
• Appetite loss.
• Fatigue.
• Fever, low-grade.
• Muscle and joint aches.
• Nausea and vomiting.
• Yellowish skin, dark yellow urine due to
jaundice.
1.0.4 Diagnosis and Test for HBV
The following test are done to identify and
monitor liver damage from hepatitis B;
• Albumin level.
• Liver function test.
• Prothrombin time.
•
•
•
•
The following test are done to help
diagnose and monitor people with
hepatitis B,
Antibody to HBsAg (anti-HBs): a positive
result means the body has either had
hepatitis B in the past or has received a
hepatitis B vaccine.
Anti body to hepatitis B core antigen (antiHBc): a positive result means the body
has recently been infected or in the past.
Hepatitis B surface antigen (HBsAg): a
positive result means that the body has an
acute infection.
Hepatitis E surface antigen (HBe Ag): a
positive result means the body has
hepatitis B infection and is more likely to
spread the infection to others through
sexual contact or sharing needles.
Patients with chronic hepatitis will need ongoing
blood test to monitor their status.
1.0.5 Treatment of HBV
Acute hepatitis needs no treatment other than
careful monitoring of the liver and other body
functions with blood test. Therefore the patient
should get the following.
• Plenty of bed rest.
• Plenty of fluids to drink.
• Eating healthy foods.
• Treatment with alpha inferno and steroids
given by injjection. This stimulates the
body’s immune system and is administered
for 16 weeks; it has no serious side effect.
• Treatment with lamuvudine, a drug taken
orally for 52 weeks. Side effects are not
serious.
• It is in rare cases that may need liver
transplant due to liver failure.
Patients with chronic hepatitis b should avoid
alcohol and should always check with the doctor
before taking over the counter medication or herbal
supplements. This even includes medication such as
acetaminophen, aspirin and ibuprofen.
1.0
Objectives
The process of diagnosing and treating hepatitis
B virus with this software is not easy but with
mutual bridge between the doctors, patients and the
computer will make things faster and more
efficient. Patients are to be under careful
supervision for a long period of time.
The key objective here is to develop an
application programme that is user friendly and can
diagnose and treat hepatitis B virus both the control
forms logically and functionally will be related
within the system and is available to monitor
patients and should be consistent.
.
2.0
Design Methodology:
The steps that has been adopted in the diagnosis
and treatment of hepatitis B are as follows:
analysis of current system, problem of the current
system, analysis of the proposed system, system
design, which are classified under Structured
Systems Analysis and Design Methodology
(SSADM).
The inference engine uses problem-solving
knowledge or methods that interacts with the user
and processes the result from the collection of rules
and data in the knowledge base. An expert-system
shell provides customizable inference engines and
knowledge base that contains rules that are of the
form “IF condition THEN action”. The condition
portion of the rule is usually a fact inputted by the
user, the action portion of the rule can include
actions that affect the outside world, activate
another rule or add a new fact to the database. It has
the capacity to acquire, store, retrieve,
communicate, process and use knowledge for the
purpose of solving problem. Figure 1 shows the
Rule-based Expert System of the proposed system.
.
Figure 1: structure of the rule based expert system and developmental process
The methods to be taken are indicated in the following steps:
Step I
A feasibility study will be carried out to acquire
knowledge about the activities of the physicians.
This study also investigates the project and can
either justify the development of the project or
shows reasons why the project should not be
continued.
Step II
Investigation and fact finding, which is a detailed
study, will be conducted so as to identify the basic
information requirements. It also involves
contributions from the end users (asthmatics and
experts) as well. They can easily pin-point the flaws
in the old manual system and suggest
improvements.
Step III
Analysis - this gives a full description of the old
(manual) system and its objective. The manual
system in this context refers to the traditional
approach to diagnosis which involves direct contact
with the medical doctor or personnel as the case
may be.
Step IV
Design is based on analysis carried out in the
previous step and information gathered in previous
steps. Interface will be created taken cognizance of
the recommendations given in the previous stages.
Coding comes in after the interface design; errors
will be debugged and then the project can be
implemented.
Step V
Training: in a situation where the project is to be
adopted, there is a need to perform some training on
the would-be users. The flowchart in figure (2)
below, shows how the various subsystems or
modules operate. The three (3) modules
(subsystems) of the system are:
i. Patient information
ii. Diagnosis
iii. Result and medication
Figure 2 Patient information and diagnosis flowchart
1
.
Figure (3) indicates the system flow diagram, in
which the database that will be used is MySQL
because it is the world’s most free and open source
software (FOSS) database. MySQL is a relational
database management system which was drawn
from the need to quickly and efficiently deploy a
data base application. Java is the programming
Language adopted for coding the application. The
data base:
• Stores users (patients) information
•
•
Retrieves patient’s information
Updates records.
Single table will be designed to handle the
various records of the patient that needed to be
stored (such as gender, age, answers to the
diagnostic questions, recommendation e.t.c ) so as
to enable the doctor retrieve the existing patient
previous records. Also, the table will be properly
decomposed so as to avoid repeating values.
Figure 3: System Flow Diagram.
3.0
Results and Discussion.
The program has several windows (frames) each
performing some certain tasks. In registering the
patient, the first window is the patient information
window which takes the information of the patient
and performs some validity check on them. It is a
separate class with only one constructor as:
public Frame1() {
initComponents();
}
When the information provided by the patient is
authenticated, it is sent to another window called
.
the diagnostic test window. A separate class is also
created for this task. The class has one constructor
thus:
public Frame2() {
initComponents();
}
public void setInformation(String name, String
state, String town, String age, String gender, String
marital, String regNo){
nameField.setText(name.toUpperCase());
noField.setText(regNo.toUpperCase());
stateField.setText(state.toUpperCase());
townField.setText(town.toUpperCase());
ageField.setText(age);
genderField.setText(gender.toUpperCase());
maritalField.setText(marital.toUpperCase());
}
Diagnosis test questions are asked as follows: and
shown the interface in figure 5.
1.
Do u have fever?
2.
Do you have loss of appetite?
3.
Do you have nausea and vomiting?
4.
Do you have fatigue?
5.
Do you have dark yellow urine?
6.
Have you received any blood
transfusion in the last 3-6 months?
7.
Do you feel pains on the right hand
side of your abdomen?
8.
Does anyone in your house or family
have or have been treated of hepatitis?
9.
Have you nursed a patient with
hepatitis recently?
10.
Do you live in overcrowded
environment?
11.
Do you share formite (cloths) or any
12.
13.
personal item like (toothbrush, razor e.t.c)?
Have you ever received injection
using any unsterilized needle or tattoo with
unsterilized instrument?
Have you had unprotected sex with
someone you suspect to have hepatitis.?
And fields for providing answer to each of
these questions are provided. When the user
clicks the submit button, the answers are
passed to another class called
Test to diagnose hepatitis thus:
1.
HBs Ag test
2.
HBe Ag test
3.
Anti-Hbc test
4.
LFT( liver function test)
5.
Urinalysis
6.
Abdominal scan
Also a separate class is created for this which has
one constructor and no main method.
public ProgramLogic(){
}
The function of this class is to analyze the
answers supplied by the user in order to determine
whether the patient has hepatitis B or not.
If–else if – else statements are constructed in
some amazing order to make correct decision. If
test 1,2,3,5 are positive, patient should be placed on
regular check up, treated and advised.
If any of test 4 and 6 are positive, patient should
be placed on constant check up, treated with
stronger medication, monitored and advised.
If test 6 is severely damaged, patient may need
a liver transplant.
.
Figure 4 Information Interface.
A separate class is also created for this and it is
also frame-based. It has only one constructor and
several methods. There is a method that accepts the
hepatitis B test result.
The frame has several panels. These include the
panel containing the patient information (i.e. name,
reg. No, age etc), the panel containing the
diagnostic question, responses of the patient,
displaying the test result and the panel containing
the appropriate recommendation, shown in figure
(5), the other interfaces are shown later in this
paper.
Figure 5: Interface for the test questions
The method that receives the hepatitis B test
result displays the result of the test on the result
panel. The recommendations are based on the
hepatitis B level as described in later in this paper..
These recommendations are read from an
external text file so that it can be updated easily by
updating the text file which resides in the user
home directory.
Conclusion:
A feasibility study was carried out through
interviews with medical experts so as to extract
expertise about hepatitis B. Doctors were
interviewed to gain insight into their expectation as
they would be part of the end-users. After taking
into consideration the facts gained from the
interviews and the questioners the output of this
investigation was analyzed and the design was
made which was successfully implemented. This
project explains and shows how possible expert
systems for the diagnosis of hepatitis B can be
implemented. Below are some benefits:
a.
It makes diagnosis faster and less
prone to errors
b.
The operation proved to be more
consistent and accurate compared to
the existing system
The project provides adequate solutions to the
problems mentioned. It integrates expert systems
into healthcare services via the creation of an expert
system for hepatitis B diagnosis and management.
If this project is fully implemented it will greatly
aid the distribution of primary health care services
around Nigeria, Africa and the globe.
The result of this project has shown that an
expert system for diagnosis and management of
hepatitis B would be of immense help to hepatitis,
non-hepatitis, medical experts and all who are
interested in gaining information about hepatitis B
and its symptoms.
This system is not meant to replace doctors but to
assist them in the quality service they render to
humanity. The diagnostic capacity of a medical
expert using this System improves only slightly
compared with his/her capacity without the aid of
the system.
1
________________________________________________________________________________
References
[1]
[2]
[3]
[4]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
Rajkishore Prasad, Kumar Rajeev Ranjan, and A.K. Sinha, (2006) “AMRAPALIKA: An
Expert system for the diagnosis of pests, diseases, disorders in Indian mango,” Knowl.-Based
Syst. 19(1): 9-21 (2006).
Shikhar Kr. Sarma, Kh. Robindro Singh, An Expert System for diagnosis of diseases in Rice
Plant, Department of Computer Science, Gauhati University Guwahati 781014: Assam, India
Ganesan V., “Decision Support System “Crop-9-DSS” for Identified Crops”, Proceedings of
World Academy of Science, Engineering and Technology Volume 12 March 2006 ISSN 13076884 PWASET Volume.
Ali.Adeli (2010), A Fuzzy Expert System for Heart Disease Diagnosis, Proceedings of
International Multiconference of Engineers and Computer Scientists, 2010 Vol I, IMECS 2010,
March 17-19, 2010, Hongkong.
Mehdi.Neshat, A Fuzzy Expert System for Heart Disease Diagnosis, Proceedings of
International Multiconference of Engineers and Computer Scientists, 2010 Vol I, IMECS 2010,
March 17-19, 2010, Hongkong.
Ahmed Rafea, “Expert System Applications: Agriculture,” Central Laboratory for
Agricultural Expert Systems, P.O.Box 100 Dokki Giza, Egypt.
P.Santosh Kumar Patra, An Expert System for Diagnosis of Human diseases, 2010
International Journal of Computer Applications (0975 – 8887) Volume 1 – No. 13.
Russell, S. and P. Norvig, 2002. Artificial Intelligence: A Modern Approach, Prentice Hall,
Second Edition.
Ahmad T. Al-Taani, An Expert System for Car Failure Diagnosis, World Academy of
Science, Engineering and Technology 12, 2005.
Fahad Shahbaz Khan, Dr. Wheat: A Web-based Expert System for Diagnosis of Diseases
and Pests in Pakistani Wheat, Proceedings of the World Congress on Engineering 2008 Vol I ,
WCE 2008, July 2 - 4, 2008, London, U.K.
Duan, Y., Edwards, J.S., and Xu, (2005) M.X. Web-based expert systems: Benefits and
challenges. Information & Management, 42 (2005), 799-811.
1
.
A Comparative Performance Analysis of Popular Internet Browsers in Current Web
Applications
Boukari Souley, Amina S. Sambo
Mathematical Sciences Programme, Abubakar Tafawa Balewa University(ATBU), Bauchi, Nigeria
bsouley2001@yahoo.com +2348069667696 +2348033512019
Abstract
As more and more services become available on the Internet, the issue of fast and secured access to online
resources gains more importance. The growth of the internet has encouraged a high number of people to
explore and take advantage of the World Wide Web (www). The window to the World Wide Web is a web
browser, hence the development of various web browsers in the market today. A comparative study of four web
browsers namely Internet Explorer, Mozilla Firefox, Opera and Netscape was carried out. A web portal
developed and fully tested was used to evaluate the performance of the four browsers. The results revealed that
Mozilla Firefox and Netscape perform best in down load time, Internet Explorer performs best in memory
usage, privacy and security, Mozilla Firefox and Netscape performs best in Page Layout, and Opera performs
best in speed and performance.
Key words: World Wide Web, Internet Browser, W3C, APIs, Milnet
Introduction
The World Wide Web (WWW) is one of the
most accessible parts of the Internet. Its ease of use
allows people to perform tasks quickly and easily.
But the creation of online shopping, banking,
search engines, corporate websites and other
personal services leads many users to pass
information that should be kept private in an
environment to which potentially everyone could
have access. Web browsers attempt to present to
user the best presentation they can offer and other
options to facilitate better services on the web.
Similarly, browsers also attempt to notify the user
when applications are downloaded and try to
execute on the user’s machine. However, the result
of various browsers based on some useful
characteristics differs.
Though Web standards do exist, different
browsers behave differently (in fact, the same
browser may behave differently depending on the
platform). Many browsers, such as Internet
Explorer, also support pre-W3C APIs and have
never added extensive support for the W3Ccompliant ones[11].
Examples of web browsers include Netscape
Navigator, Mozilla Firefox, Internet Explorer,
Opera, Lynx, Enigma and so on. For browser
compatibility, there are basic ways one can make a
.
Web application extensible in order to add new
browser support later.
In developing web applications, consideration
must be given to possible browser differences;
hence the developer should be informed about
them. Following those guidelines not only allow
your web applications to work in other browsers,
but also on other platforms.
Web browsers are now an essential part of our
daily lives. Many people use them to access e-mail,
perform research, buy products and do other
errands. Because web browsers are used for so
many tasks, there are built- in functions to perform
those tasks as well as to protect users from
malicious content on the World Wide Web.
Generally, browsers react and display
information differently and cross browser
compatibility should be considered during web
design. The World Wide Web contains millions of
web pages with a variety of different types of
content.
The lack of awareness of available web browser
and their characteristics, discourages many
Nigerians from having access and enjoy faster
access to vast mount of free information available
on the internet.
A high volume of commercial activities occur on
the internet not to mention free online resources
that will enable research and empowering us with
knowledge. Internet access and awareness are
currently available in most part of Nigeria. Though
bandwidth is restricted, Surfers should be aware of
the browser options they have and select them
based on their most important criteria.
For developers, the availability of a tool that
performs the comparison helps to speed up the
development time, assists them to become aware of
the short comings, features and capabilities of each
of the four web browsers. This reduces the time it
takes to track down bugs in the web applications
hence improves development time and application
performance.
This study of web browsers becomes imperative
to help serve as a educational and decision making
tool for users. This study exploits on and explains
the web browser characteristics of four web
browsers from the various available web browsers.
Related Work
Brief history of the internet
The internet started as a project called the
Advanced Research Projects Administration
Network (ARPANET). It was meant to be both, an
experiment in reliable networking and to link the
American Department of Defense (DOD) with
military contractors and universities doing military
funded research. It was launched in 1969 and
started connecting three supercomputers in
California, U.S.A and one in Utah, USA. The
success of ARPANET caused many universities to
desire to be connected to the network [1]
This caused its growth and eventually it became
hard to manage. It then split into two parts.
MILNET, which catered for military sites only, and
a new smaller ARPANET for non-military sites.
Around 1980, the American National Sciences
Foundation (NSF) decided to create five
supercomputing centers for research use. The NSF
created its own network NSFNET.
By, 1990
almost all ARPANET traffic had been moved to
NSFNET. In 1994, several large, commercial
networks had been created within what is now
called the internet. These networks grew beyond
the borders of the United State and are today
.
connected to almost all countries of the world.
Other networks in other countries also linked up to
create the internet, as it is known today [2].
In [3] internet is defined as a sprawling collection
of computer networks that span the globe,
connecting government, military educational and
commercial institutions, as well as private citizens
to a wide range of computer services, resources and
information. A set of network conventions and
common tools are employed to give the appearance
of a single large network even though the
computers that are linked together use many
different hardware and software platforms.
A non technical definition of the internet as seen
by an average internet user would be,
“A virtual world assessed through computers and
other devices. This virtual world modeled after the
real world. This world now informally called
‘cyberspace’, has been evolving over the years and
has continued to evolve..
Internet Services.
The internet offers many services to its users.
The most popular services include E-mail, World
Wide Web (WWW) and File Transfer protocol
(FTP). Other less popular services offer access to
other types of internet resources. These include
Telnet, Finger and so on.
E-Mail
E-mail is the most popular internet service and it
provides reliable means of communication
worldwide. It consists of information, usually text
that is electronically transferred over telephone
lines, fiber optic cables and satellite links, from a
single sender to one or more recipients. Two pieces
of information are needed to send e-mail messages
over the internet - the recipients’ user identification
and the computer name to which the mail is to be
sent. Simple Mail Transfer Protocol (SMTP) and
Post Office Protocol (POP) are used to standardize
the transmission format [7].
File Transfer Protocol (FTP):
This is a means of accessing files on a remote
computer system, also called an ftp site. These files
are stored in a tree like structures of directories. A
connection is established with the computer system,
the desired files are located and copied or
downloaded onto the Users’ hard disk. This allows
information to be populated to the internet [9].
World Wide Web:
The WWW is the fastest growing internet
service and it is treated as the future of internet
navigational tools. It is a multimedia and hypertext
system that allows pictures, video clips and sound
to be included in text pages. The pages also have
links to other pages that are displayed when the
links are selected using a pointing device or
keyboard. The WWW consists of documents called
web pages and a collection of web pages on a
particular subject matter stored on a single system
from a website. The WWW uses the Hypertext
Transfer Protocol (HTTP) to transmit web pages.
Documents are viewed using software applications
called web browsers[5].
Internet Browser and Characteristics
Mozilla Firefox : Mozilla Firefox (originally
known as "Phoenix" and briefly as "Mozilla
Firebird") is a free, cross-platform, graphical web
browser developed by the Mozilla Foundation and
hundreds of volunteers.
.
The features of Mozilla Firefox distinguish it
from other web browsers such as Internet
Explorer, and are subject to both rave reviews and
harsh criticisms. It lacks many features found in
other browsers, in an effort to combat interface
bloat and to allow the browser to be shipped as a
small, pared-down core easily customizable to meet
individual users' needs. Instead of providing all
features in the standard distribution, Mozilla
Firefox relies on the extension system to allow
users to modify the browser according to their
requirements.
Internet Explorer: Internet Explorer (IE) is the
popular Web browser created and distributed by
Microsoft. IE was first released in 1995, and IE has
been the most popular Web browser since 1999 [6].
Internet Explorer has been designed to view the
broadest range of web pages without major
problems. During the heydays of the historic
browser wars, Internet Explorer embraced Netscape
by supporting many of the progressive features of
the time. For a long period after the introduction of
version six, there was no further development on
the browser. Major development on the browser
restarted in 2004 for Windows XP SP2 and
continues in IE7 [4].
Netscape Browser: Netscape Browser is the name
of a proprietary Windows web browser published
by American Online, but developed by Mercurial
Communications. It is a continuation in name of the
Netscape series of browsers, originally produced by
the
defunct
Netscape
Communications
Corporation[10].
While Netscape Browser's version numbers start
at 8, it is based on Mozilla Firefox, whereas
Netscape 6 and 7 were based on Mozilla
Application Suite, itself a complete rewrite of the
codebase developed in versions 1 through 4
Netscape Navigator and Netscape Communicator.
As with other recent versions, it incorporates
support for AOL Instant Messenger, and other
AOL-related features [3].
Perhaps the most noteworthy feature introduced
in Netscape Browser is the ability to use either of
two layout engines to render websites — either
Internet Explorer's Trident or the Gecko engine
used by Mozilla and its derivatives. This is used as
part of the browser's "Site Controls" system, which
allows security settings to be altered on a per-site
basis, and is also touted as a defense against
phishing and similar attacks, with both blacklists
and white lists built in and automatically updated
regularly. This system decides whether a site is
"trusted" or "suspect", while only white listed sites
use the Trident engine by default [8].
Other features highlighted by AOL's publicity
include improved tabbed browsing capabilities, a
toolbar system called the "MultiBar," which
includes up to ten toolbars in the space of one, and
extra support for "Live Content", such as RSS
feeds. In keeping with the security emphasis, a new
secure form information and password management
system, known as a "PassCard," which saves
usernames and passwords for individual sites, is
also included[11].
Opera: Opera is a cross-platform web browser and
Internet suite which handles common internetrelated tasks including visiting web sites, sending
and receiving e-mail messages, managing contacts,
chatting online and displaying Widgets. Opera's
lightweight mobile web browser Opera Mini and
.
most current versions of its desktop application are
offered free of charge[12].
Opera is proprietary software developed by
Opera Software based in Oslo, Norway. It runs on a
variety of operating systems including many
versions of Microsoft Windows, Mac OS X, Linux,
FreeBSD and Solaris. It is also used in mobile
phones, smartphones, Personal Digital Assistants,
game consoles and interactive televisions.
Technology from Opera is also licensed by other
companies for use in such products as Adobe
Creative Suite[4].
Methodology
To successfully carry out an analytical
comparison on these major web browsers, three
major points have to be brought into consideration,
namely - Performance, Usability and Security
Putting into consideration these three points, the
comparison will be based on the following Criteria.
• Download time: How long it takes to load a
web page
• Memory usage: The amount of memory
each web browsers use
• Page Layout/Image Display: How each
handle / displays a web page
• Accessibility: How each complies to the
international accessibility Standards
• Privacy and Security1
• Speed And Performance: These includes:
o Rendering Cascading Style Sheet
(CSS)
o Rendering table
o Script speedEase of Setup: How
easy it is to install on a users
computer
Another step is determining what others tried
doing to get around the issue of unusable user
interfaces. Studies of existing browser reveal that
while the user may really want to use the most
popular internet browser other browsers have better
features but lack of awareness is restricting
them[9].
A website that will allow for the comparison to
be carried out was designed using the concept of
web portal, JavaScript
and Macromedia
Dreamweaver as the development tool and HTML
for the front end user interface.
Download Time
To make a proper comparison between both
browers of how long it takes to download a
specified web page.
Memory Usage
To effectively compare how much memory each
web browser uses could cover two aspects.
a.
How much memory is used when
multiple windows are opened
b.
How much memory is used when
multiple tabs are opened
To achieve each, the windows task manager would
be used. The Task Manager calculates and displays
how much memory each currently running
applications is consuming.
Page Layout / Image Display
To achieve this test, a web page is created that
contains common HTML elements that are
common to websites. The web page would also
contain all the various image file formats in the
.
industry today. This page is now viewed using the
various web browsers and a comparison is made
against each other.
a.
b.
The key points here would be:
a.
How each layout and displays the web
page
b.
How each is able to handle, support
and display the various image file formats in
the industry today.
The results are compared and analysis made based
on the outcome.
Accessibility
Web accessibility means that people with
disabilities can perceive, understand, navigate, and
interact with the Web, and that they can contribute
to the Web. Web accessibility also benefits others,
including older people with changing abilities due
to aging[8]
This section would be based on already
concluded test carried out by international
communities that major on accessibility issue. The
objective here is to determine to what extent each
web browser implements the accessibility standards
set by the international community.
The process would be as follows:
1.
List out the major associability standards
set by the international community
2.
Determine to what extent each web
Browser implements each.
Privacy and Security
This would be to determine how secure each web
browser is. To achieve this, a web page would be
created that contains the following elements.
c.
Applets, Scripts and ActiveX
Object
A link to download an external
document
A link to execute an external
program.
The essence of this test is to see how each of
these web browsers secures the users from external
third party programs from running on the host
system.
Speed and Performance
The goal of this test is to determine how fast
each of these web browses handles and executes
various programs or instructions. The test would be
carried out on individual web browser premises.
1.
Rendering Cascading Style Sheets
(CSS)
CSS which stands for Cascading Style Sheets, is
the technology that make web pages look the way
they look. It is a set of instructions given to a web
browser to tell it how the web page is to be
displayed, e.g. colors, font sizes, backgrounds etc.
The essence of this test is to find out how
effectively the web browser executes sets of this
instruction within each web browser[12]
Rendering table and Script Speed
Tables are a common feature in most web pages.
This test procedure is similar to the earlier
mentioned process. We load a web page that
contains lots of table elements and determine how
well and how long it takes each web browser to
render the tables, then we compare results.
.
This test would determine the length of time it
take for each web browser to successfully complete
the execution of a certain set(s) of commands.
Scripts are a vital part of the functionality and
interactivity of many web pages. The essence of
this test is to see how reliable each web browsers in
the execution of scripts.
Ease Of Setup
This test would require the installation of each of
the web browsers in questions and determine how
difficult or easy it was to successfully install the
software on the user’s machine.
Results can be taken from already carried out
studies to save the user (supervisor) the hassles of
having to uninstall and reinstall the web browser.
CSS 3
Run Tests.
Table 1: Scoring and Rating
Test
Browser Info
CSS 1
CSS 2
Score
5 Points
3 points (1
each)
4 points (1
each)
Results and Findings
A web browser analysis tool was designed, fully
implemented and tested to enable users make
informed decisions about the use, installation and
recommendation of the major web browsers
considered in work. The tool was designed and
implemented using industry standard technologies
such as:
1. JavaScript Scripting Language
2. Hypertext Markup language (HTML)
3. Document Object Model (DOM)
4. Cascading Style Sheet (CSS)
The major Criteria used for comparison are:
Download Time,Memory Usage, Page/Image
Layout, Accessibility, Privacy and Security, Speed
and Performance
Test
ActiveXObject
Executable
Score
1 Point
1 Point
Download
1 Point
The RunTests first set the result display area to a
default value of 0 (Zero), it then proceeds to call 8
Graphics
Card Layout
2 Points (1
each)
4 Points (1
each)
5 Points
Accessibility
3
Points(1
each)
(eight) other functions (TryCatch fucntion, Layer
Movement, Random number engine, Math Engine,
Dom Speed, Array Functions, String Function,
Ajax Declaration)
.
Results
After testing on the various platforms, the Results obtained were summarized in Table 2 below:
Table 2: Result of each browser based on the criteria
Download
Time
Memory
Usage
Page Layout
Internet Explorer
Mozilla Firefox
Opera
Netscape
17 seconds
4 seconds
15 seconds
4 seconds
6kb
523kb
726kb
233kb
12 points
16 points
15 points
16 points
2 points
2 points
2 points
3 points
3 points
3 points
Privacy
/ 3 points
Security
Accessibility 1 point
CSS – 60 ms CSS – 150 ms
Speed And CSS – 80 ms CSS – 241 ms
Performance SCRIPT – 3876ms SCRIPT – 4917 SCRIPT – 2233 SCRIPT – 0 (*)
MS
ms
Discussion
Based on the results in the test run as shown in
Table 2 above, the following observations have
been made:
Download Time: Internet Explorer performs best,
followed by Opera, Netscape and Mozilla Firefox
presents the same performance.
Memory Usage: Opera uses the highest amount of
memory available to the system, followed by
Mozilla Firefox, Internet Explorer, in this test
preformed best in the utilization of memory
Page Layout: Mozilla Mozilla Firefox preformed
best in the presentation and implementation of
graphics and CSS style definitions followed by
Netscape, then Opera. Internet Explorer performed
the least in this tests.
Privacy and Security: Internet Explorer gave the
highest notification alerts to third party activities
within the web browser. Others performed the
same.
Accessibility: All the web browsers tested
performed equally on this test except for Internet
Explorer, scoring only a point.
Speed and Performance: The speed test was
carried out on two premises.
.
1.
CSS Speed Test: In this category, Opera
showed a higher performance in executing CSS
definitions, followed closely by Internet Explorer.
Mozilla Firefox and Netscape, performed slower.
2.
Script Speed Test: Opera also, in this test,
performed better in executing the various
JavaScript commands that where sent to the
scripting engine. It is followed by Internet Explorer
in performance then Mozilla Firefox. Netscape
could not successfully execute the process, hence
failed in the test.
Conclusion
A study of four internet browsers was carried
out. A portal was designed that serves as a decision
.
or intelligent tool for analyzing different web
browsers. Depending on the browser on your
system, the capacity of the system in terms of
memory, speed and brand, results will be obtained
for the criteria of comparison of browsers. This
intelligent decision analysis tool will enlighten
people on browser technology, possibility and
finally will enable people to know which web
browsers to use based on their various purposes.
From the test results of the four internet browsers, a
researcher would find internet explorer best because
of its speed of download and security feature while
a graphic programmer will find Mozilla Firefox
more favorable because of its high performance of
page
layout
__________________________________________________________________________
References
[1]
[2]
[3]
[4].
[5]
[6].
[7]
[8]
[9]
[10]
[11]
[12]
Vigil,B. (2002), Tech Target Security Media website.
www.searchsecurity.techtarget.com
Oppliger, R., (2000) Security Technologies for the World Wide Web. Artech House,
Boston
Raskin, J. (2000) The Humane Interface: New Directions for Designing Interactive
Systems. Addison Wesley Longman, Inc, London
Tiwana, A..(1999) Web Security. Digital Press, Boston.
Brian B.. (2004)Browser Showdown Mozilla Firefox vs. Internet Explorer
Rankin, (1997) Web browsers Addison Wesley Longman, Inc, London
http://www.agnitum.com/news/securityinsight/december2005issue.php
http://www.rnib.org.uk/xpedio/groups/public/documents/PublicWebsite
http://www.aadmm.de/en/month.htm
http://www.agnitum.com/news/securityinsight/december2005issue.php
http://www.us-cert.gov/cas/tips
http://en.wikipedia.org/wiki/Comparison_of_web_browsers
.
Adjusting for the Incidence of Measurement Errors in Multilevel Models Using
Bootstrapping and Gibbs Sampling Techniques
Imande, M.T* and Bamiduro, T.A+
*Department of Mathematics and Computer Science, E-mail: imandemichaelt@yahoo.com
Benue State University, P.M.B. 102119, Makurdi , Nigeria.
+
Department of Statistics, University of Ibadan, Nigeria.
Abstract
In the face of seeming dearth of objective methods of estimating measurement error variance and realistically
adjusting for the incidence of measurement errors in multilevel models, researchers often indulge in the
traditional approach of arbitrary choice of measurement error variance and this has the potential of giving
misleading inferences. This paper employs bootstrapping and Gibbs Sampling techniques to systematically
estimate measurement error variance of selected error-prone predictor variables and adjusts for measurement
errors in 2 and 4 level model frameworks. Five illustrative data sets, partly supplemented through simulation,
were drawn from an educational environment giving rise to the multilevel structures needed. Adjusting for the
incidence of measurement errors using these techniques generally revealed coefficient estimates of error-prone
predictors to have increased numerical value, increased standard error, reduced overall model deviance and
reduced coefficient of variation. The techniques, however, performed better for error-prone predictor(s) having
random coefficients. It is opined that the bootstrapping and Gibbs Sampling techniques for adjusting for the
incidence of measurement errors in multilevel models is systematic and realistic enough to employ in respect of
error-prone predictors that have random coefficients and adjustments that are meaningful should be appraised
taking into cognizance changes in the coefficient of variation alongside other traditionally expected changes
that should follow measurement error adjustments.
Key words: Multilevel models, Measurement error adjustment, Coefficient of variation,
Predictor variables, Bootstrapping, Gibbs sampling.
_______________________________________________________________________________
1.0 Introduction
In many of the variables used in the physical,
biological,
social
and
medical
science,
measurement errors are found. The errors are
essentially random or systematic. Both types of
errors could be problematic in statistical inference.
In fixed effects models such as linear
and
generalized linear models, the incidence and effects
of measurement errors on the response and
explanatory variables has been well documented in
the literature [4], [9], [2], [8], [1], [12], [4].
Generally, the consequences of ignoring
measurement errors for independent observations
.
and response values are well understood in linear
models.
The efficacy of
mixed models such as
multilevel linear models is also adversely affected
by a failure to properly account for measurement
errors in their formulation and estimation. In
particular, the behaviour of biases associated with
measurement error in covariates or the response for
multilevel hierarchical linear models is, up to date,
not well known and can be complex [7]. In
handling the incidence of measurement errors in
multilevel modeling methodology, one of the
daunting challenges that often confront researchers
is that of estimating realistically measurement
error variances and reliabilities of error-prone
variables in a multilevel model. Most of the current
techniques for estimating
measurement error
variance are, in general deficient; there is inability
to sufficiently
justify
independence of
measurement errors and the so called
unidimensionality assumption as required in
educational
mental testing; accuracy and
consistency of the estimates of the measurement
error variance could not be guaranteed [3]. The
method of instrumental variables strongly
recommended for certain situations as in mental
testing (see [3] ) requires, however, that several
different instrumental variables be considered for
comparison. There is also often the difficulty of
establishing that measurement errors are
independent of instrumental variables [11]. Some
other researchers often simply assume measurement
error variance and reliability values for error-prone
variables in the multilevel models at the risk of
obtaining unrealistic estimates. This paper employs
bootstrapping and Gibbs sampling techniques to
realistically estimate measurement error variances
of selected error-prone explanatory variables and
adjusts for the incidence of these errors giving rise
to more adequate multilevel models.
2.0 Methodology
2.1
Data Structure
The illustrative data employed was drawn from
an educational environment. There were five data
sets( Data 1-5) utilized. Data 1-3 were derived
from 50 randomly selected secondary schools in
Benue State of Nigeria while Data 4 and 5 included
data supplemented by simulated values. Data 1
constituted a 4-level data structure in which there
were 9,999 level 1 units (here students), 450 level
2 units (here subjects or subject groups), 150 level
3 units (here classes) and 50 level 4 units (here
schools). The clustering was such that for any
original sample nj ( 20 ≤ nj ≤ 30) of the students
from each school j, the nj was “replicated” into 9
clusters giving rise to 9nj level 1 units for school j (j
= 1,2, …,50). In other words, the same nj students
in school j were mirrored in 9 clusters or groups
and, in particular, for each school j, we had 9nj
level 1 units nested in 9 level 2 units that were
further nested in 3 level 3 units . Data 2 also
constituted a 4-level data structure but here there
were 6,666 level 1 units(students), 300 level 2
units( subjects or subject groups), 150 level 3 units(
classes) and 50 level 4 units( schools); in this
dataset, the seeming confounding characteristics in
Data 1 were reduced by removing the level 2 unit or
cluster relating to Common Entrance (CE) and
variables based on it. Data 3 is a 2-level data
structure with students nested in schools; any
sample drawn in a school constituted a “statistical
cohort” of students from whom Mathematics (M)
and Science and Technology (ST) scores in JSS1,
.
JSSCE and SSCE/WAEC between 2002 and 2008
were captured. Data 3 had 1,111 level 1 units and
50 level 2 units. Additional levels 1 and 2 units
were further generated via simulation to supplement
needed data for further exploration. These gave rise
to Data 4 (having 2,222 level 1 units with same 50
level 2 units) and Data 5( having 4022 level 1 units
and 110 level 2 units).
2.2 Description of Variables
Variable name
Description of Variable
Navgstemij
Ncescoreij
Normscoreij
Navg1steml
Navgcel
Navg2steml
Navg3stemj
Navgsubj
Navginclsk
Navgscoreij
NJS1avgij
NJCEavgij
Schstatusl
Schsysteml
Schgenderl
Data set where
used
STM score ; a level 1 response 1, 2
Student’s Final
variable.
Student’s entrance score; a level 1 predictor variable
JSS1 school STM score student’s subject score per
class; a level 1 predictor variable.
Final School STM score; a level 4 predictor variable
School common entrance score; a level 4 predictor
variable.
JSSCE school STM score ; a level 4 predictor
variable.
Final School STM score; a level 4 predictor variable
Score per subject; a level 2 predictor variable.
Score in class ; a level 3 predictor variable.
STM score per student in all classes; a level 1
response variable.
STM score per student in JSS1 subjects; a level 1
predictor variable
STM score per student in JSSCE subjects; a level 1
predictor variable.
1
1, 2
1, 2
1
1, 2
1, 2
1, 2
1, 2
3-5
3-5
3-5
school status(i.e whether school is owned as private 3-5
or public); a categorical predictor variable.
school system; it is a categorical predictor variable 1-5
with the systems categorized into “Boardsytem”, “
Daysystem” or “Bothsystem”.
School gender ; it is categorical predictor variable 1-5
with school gender categorized into Boys
school(Boysch), Girls school (Girlsch) or Mixed(
Mixedsch).
.
Nrsqindexl
PSStatusl
Labavl
School staff quality index (an indication of academic 1-5
staff quality or strength in any particular school. This
is estimated by dividing the total number of qualified
academic staff by the entire estimated student
population in the school; it is a predictor variable.
An indication of Electric Power Supply status in a 3, 4
school ; it is a categorical predictor variable
categorized into school generator, PHCN, Both or
one.
An indication of the availability of Science 3
Laboratories in a school; it is a categorical predictor
variable categorized into “no science lab”, “ one
science lab “ or “ two or more science labs” .
3 Multilevel Models and Measurement Errors
A k-level model may be expressed in the
compact form:
Y = Xγ + ZU + Z(1)e
(2.1)
where, Y is a column vector of true unobservable
responses each assumed continuous.
Z = [Z(k) , Z(k-1), …, Z(2)]
and
U′ = [u(k) u(k-1) ,…,u(2)] .
The Z(k)’s are block diagonal matrices having
diagonal elements as Zj(k) ( j= 1,2,…,mk) while
u(k), X and γ are column matrices with elements,
respectively, uj(k) , Xj
(j=1,2,…,mk),and γh0
(h=0,1,…,p).
We assume that Z(1)e and U are normally
distributed with zero mean and we, symbolically,
write:
Z(1)e = r ~ N (O, σ2 Ï*)
…
(2.2)
and
U ~ N (O, T*)
…
(2.3)
where Ï* and T* are appropriate block diagonal
matrices comprising, respectively, the blocks of
unit matrices and blocks of variance-covariance
matrices of the residual vectors associated with the
k-level model (that is the residual contributions
from the levels 2, 3, …, k in the k-level model).
We infer from (2.1),(2.2) and (2.3) that Y is
normally distributed with E(Y) = Xγ and variance~ ~
covariance matrix, Vk = V =E[ E E ’] = ∑ {Vk ( l ) }
l
~
,where E = ZU + Z(1)e . The notation Vk here
referring to the covariance( or variance-covariance)
matrix associated with the response vector for the
k-level model and Vk(l)( l = 1,2,…k), respectively,
denote the contributions to the covariance matrix of
the response vector from levels k, k-1,…,1 in a klevel model. The level 1 residuals are assumed to be
independent across level 1 units. Similarly, levels 2,
3,…,k residuals are assumed to be independent
across levels 2 ,3,…, k units respectively. It should
be noted also that Vk is a block diagonal matrix
.
with block diagonal elements Vk(l)( l = 1,2,…k)
and each of these elements is also block diagonal
comprising blocks in their composition.
If the collection or measurement of explanatory or
response variables incorporated in (2.1)
are
susceptible to errors then the estimated coefficient
parameters will be asymptotically biased and
consequently incorrect inferences can result in
explanatory variables and responses takes the form.
~
Y = ~y + q~
~
Y = [Y11 Y21...Yn1…Y1j Y2j...Ynjj…Y1J Y2J… YnJJ]'
~y = [y y ...y …y y ...y …y y … y ]'
11 21
n1
1j 2j
njj
1J 2J
nJJ
~
q = [q11 q21...qn1…q1j q2j...qnjj…q1J q2J… qnJJ]'
Xij = [x0ij x1ij… xhij.. xpij] and x0ij=1
respect of the relevance or otherwise of some model
variables. In practice explanatory or response
variables utilized to fit models in social or
educational environments are subject to some
degree of measurement error.
A basic model for measurement errors in a 2level continuous response linear model for p
(2.4)
In respect of the explanatory variables or predictors, we have
~
~
x + m
(2.5)
X = ~
where
~
X = [X1 X2,.. XJ]'
~
x =[x0 x1… xh.. xp]', x0 = a column of ones.
~ = [m m …m …m ]'
m
1
2
h
p
With
xh =[xh11 xh21...xhn11…xh1j xh2j1,...,xh2jJ… xhnJJ]'
mh =[mh11 mh21...mhn11…mh1j mh2j1,...,mh2jJ… mhnJJ]'
and for each j we can write
mhj =[mh1j mh2j...mhnjj]'
xhj =[xh1j xh2j...xhnjj]'
~ and q~ are
The measurement error vectors m
assumed independent and normally distributed with
zero mean vectors. The measurement error models
reflected by (2.4) and (2,5) can be analogously
expressed in matrix form for any k-level model.
The concern of researchers and statisticians is to
seek ways of adjusting for the incidence of these
measurement errors and to do this entails a an
estimation of ( or the use of known value(s) of )
measurement error variances of perceived errorprone variables and there after use the estimated
values to modify the affected model and estimate
same.
Assuming the variables measured with error do
not have random coefficients, then following
.
Goldstein( 2003),
the ME corrected fixed
coefficients estimate for any k- level model is
-
- ... -
(2.6)
where
+ T1 + T2 + … Tk ]
(2.8)
Where T1= ⊕{βˆ ' Ω 1ijm βˆ ) and Tk
= ∑(
,
k = 1,2,…
are correction matrices for measurement errors,
Ω kαm is the covariance matrix of measurement
errors for the αth level k block, V α is the αth
block of V, the variance-covariance matrix of
residuals in the k-level model.
For the random components, based upon the model
with observed variables, we write the residual for a
unit in a level k as
(k )
= Zl
ij
=⊕
α
α
ij..l
not have random coefficients, is obtained at each
iteration as
u l(k ) + ... + Z k(3) u k(3) + Z (j 2) u (j2) +
Z ijk(1)..l eijk..l +qij..l - m'β
(2.7)
. The estimation of the variance or variancecovariance
components(i.e
the
random
components) are all estimated iteratively and, for a
k-level model, the measurement error corrected
estimate of these components, assuming the
coefficients of the variables measured with error do
for k ≥ 2.
We note that
Ω1ij is the covariance of
measurement errors for the ijth measurements of
σ 2 ij..lq is the measurement error
level 1 while
variance for the ij..lth response measurement.
If the coefficients of the variables measured with
error have random coefficients then the formulae in
~ (2.6) and (2.8) do not apply and in particular m
'V
~ has measurement errors in all its components
1m
and, following the suggestions made by
Woodhouse [13], Moment-based techniques are not
appropriate but rather the Bayesian technique of
Gibbs sampling( an MCMC technique) is
employed. Some of the selected predictor variables
perceived error-prone in this paper have random
coefficients and so Gibbs sampling technique rather
than moment-based technique shall be employed to
adjust for the incidence of these errors and, for the
estimation of measurement error variances and
reliabilities of the error-prone variables,
the
bootstrapping technique shall be employed.
.
2.4 The Multilevel Models Examined
The multilevel models formulated in respect of each of the data sets (1-5) are respectively given by (2.9),
(2.10), (2.11), (2.12) and (2.13) below.
Navgstemijkl = β0i + β1 j ( Normscore − m( Subject) )ijkl + β 2l ( Ncescore − m( Subject ))ijkl + β3 ( Navg 3stem − gm)l + β 4 ( Navgce − gm)l +
β5 DaySysteml + β 6 BothSysteml + β 7Girlschl + β8Mixedschl + β9 (Nrsqindex - gm)l + eijkl
β 0 l = β 0 + f 0l
β ij = β1 + u1 jkl
β 2 l = β 2 + f 2l
σ 2f 0

 f 0l 
~
N
(
0
,
Ω
)
:
Ω
=


f
f
f 
2
σ f 02 σ f 2 
 2l 
u 0 jkl ~ N 0, σ u20
(
(
eijkl ~ N 0, σ e2
)
)
(2.9)
Navgstem ijkl = β 0l + β 1l ( Normscore − m ( Subject )) ijkl + β 2 ( Nrsqindex − gm ) l
+ β 3 ( Navg 3stem − gm ) l + β 4 Daysystm l + β 5 Bothsystm l + β 6 Schstatus _ 1l + eijkl
β 0l = β 0 + f 0l
β 1l = β 1 + f 1l
(
eijkl ~ N 0, σ e2
)
σ 2f 0

 f 0l 
~
N
(
0
,
Ω
)
:
Ω
=


f
f
f 
2
σ f 01 σ f 1 
 1l 
(2.10)
.
Navgscoreij = β 0 j + β 1 j ( NJS1avg − m( School ) ij + β 2 j ( NJCEavg − m( School ) ij
+ β 3 ( Navg 3stem − gm) + β 4 Schstatus _ 1ij + β 5 Girlsch j + β 6 Mixedsch j + β 7 Labav _ 1 j
+ β 8 Psstatus _ 2 j + β 9 Psstatus _ 3 j + β 10 Psstatus _ 4 j + β 11 ( Nrsqindex − gm) j
+ β 12 Daysytm j + β 13 Bothsystm j + eij
with
β 0 j = β 0 + u0 j
β1 j = β 1 + u1 j
β 2 j = β 2 + u2 j
and
σ uo2
u 0 j 

 
2
u1 j  ~ N(0, Ω u ) : Ω u = σ uo1 σ u1
σ
u 
2
 2j
 uo 2 σ u12 σ u 2





e ij ~ N(0, σ e2 ) .
(2.11)
Navgscoreij = β 0 j + β 1 j ( NJS1avg − m( SCHOOL)) tij + β 2 j ( NJCEavg − m( SCHOOL)) ij
+ β 3 ( Navg 3stem − gm) j + β 4 Schstatus _ 1ij + β 5 Girlsch j + β 6 Mixedsch j + β 7 ( Nrsqindex − gm) tj +
β 8 Labav _ 1 j + β 9 Daysystm j + β 10 Bothsystm j + β 11 Psstatus _ 2 j + β 12 Psstatus _ 3 j + β 13 Psstatus _ 4 j + eij
β 0 j = β 0 + u0 j
β1 j = β 1 + u1 j
β 2 j = β 2 + u2 j
σ uo2
u 0 j 

 
2
u1 j  ~ N(0, Ω u ) : Ω u = σ uo1 σ u1
σ
u 
2
 2j
 uo 2 σ u12 σ u 2
e ij ~ N(0, σ e2 ) .





( 2.12)
.
Navgscoreij = β 0 j + β 1 j (NJS1avg - m(SCHOOL)) ij + β 2 j ( NJCEavg − m( SCHOOL)) ij
+ β 3 (Navg3stem - gm) j + β 4 Daysytm j + β 5 Bothsystm j + β 6 (NJS1avg - m(SCHOOL)).Daysystm ij
+ β 7 ( NJS1avg − m( SCHOOL)).Bothsystmij + β 8 ( Nrsqindex − gm) j + eij
β 0 j = β 0 + u0 j
β1 j = β 1 + u1 j
β 2 j = β 2 + u2 j
σ uo2
u 0 j 

 
2
u1 j  ~ N(0, Ω u ) : Ω u = σ uo1 σ u1
σ
u 
2
 2j 
 uo 2 σ u12 σ u 2





[e ] ~ N(0, Ω )
0ij
e
2.5.1 The Measurement Error
AdjustmentApproach Using Bootstrapping
and Gibbs Sampling Techniques.
The approach essentially entails re-sampling
repeatedly from each of the clusters or subgroups in
a data structure to estimate the variance of the
error-prone predictor variable, its measurement
error variance, reliability and ultimately adjusting
for the incidence of measurement errors and reestimating the k-level model accordingly. The steps
are:
(i)
From each group (or subgroup) of the
multilevel model obtain an estimate of the
Explanatory variable mean , X • j , based on
sample sizes of at least 30 in each group.
(2.13)
(ii) Average these X • j’s (using arithmetic mean)
across the entire groups to obtain a value, say X ∗ .
(iii)
Estimate the measurement error
(ME) variance, σ2hm, as the mean of the squares of
deviations of X • j’s from X ∗ .
(iv) Estimate σ2hX as in the first paradigm
approach. Estimate Rh accordingly.
(v) Use the values σ2hm and σ2hX to adjust for
measurement error in the variable (s) of interest
and hence re-estimate
the k-level model
accordingly via Gibbs sampling ; a Markov Chain
Monte Carlo(MCMC) method.
(vi) Check for possible attenuation and/or
inconsistency of the estimated multilevel
parameters
(vii) If there is attenuation (reduced or no
.
increase
in
predictive
power
of
corresponding
predictor)
and/or
inconsistency of the estimated multilevel
parameters then repeat steps (i) to (vi) ,
possibly increasing re-sampling size per
cluster and/or
increasing number
of
samples.
3.0
Analysis and Discussion
Four issues were addressed in the analysis of the
multilevel models associated with each of the five
data sets:
(a)
estimation of
the measurement error
variances and reliabilities of STM score per
student in JSS1 subjects( NJS1avg)
or its
proxies( such as Normscore variable for Data 1
and 2) as well as the School staff quality
index(Nrsqindex) predictor variables.
Table 1: Estimated Variances, Measurement Error
(M.E) Variances and Reliabilities in respect of the
(b) coefficient estimates of the perceived errorprone predictors ( and their standard errors) prior
to adjustments in measurement error.
(c) coefficient estimates of the perceived error-prone
predictors ( and their standard errors) following
adjustments in measurement error.
(d) examination of coefficient of variation values
of
coefficient estimates of the error-prone
predictors.
Bootstrapping with a minimum of 2000
replicates for each of the NJS1avg and Nrsqindex
variables in each of the data sets and following
steps (i) to (iv) we obtain measurement error
variance and reliability values as reflected in table 1
below.
‘student’s subject score per class’
Predictor
Variable or their proxies in various datasets.
Table 1: Estimated Variances, Measurement Error
Data
Variable
Variance M.E Variance Reliability
1
NJS1avg** 0.44393
0.222942
0.666872
2
NJS1avg* 0.541682 0.250571
0.683724
3
NJS1avg
0.735439 0.255298
0.742315
4
NJS1avg
0.674021 0.253647
0.726576
5
NJS1avg
0.822923 0.635357
0.564311
NJS1avg** and NJS1avg* actually refer to the
Normscore variables used for Data 1 and 2 and
are realistic proxies of the NJS1avg variable as
they are already associated with the JSS1 scores.
We find that for Data 2, 3 and 4 the variable
NJS1avg indicates a reasonably constant
measurement error variance; an average of 0.25.
In Data 1, the NJS1avg variable gave measurement
error variance estimate slightly lower (i.e 0.22)
than what obtained in Data 2, 3 and 4 but the
.
average for the Data 1-4 measurement error
variance of the NJS1avg variable is still 0.25. The
rather high measurement error variance estimate for
the NJS1avg variable (here 0.64) for Data 5 may
be attributable to weaknesses associated with the
normal probability distribution model and the
accompanying assumptions that were employed to
simulate additional units for levels 1 and 2. The
near-absence within group variation in respect of
School staff quality index (Nrsqindex) predictor
variable may have probably accounted for the high
measurement error variance estimate of 0.85
associated with it.
Using
iterative
generalized
least
squares(IGRLS) that is implemented in MLWiN
package 2.20 [10] to estimating models (2.9) -
(2.13) prior to measurement error adjustments, we
obtain coefficient estimates( with standard errors)
of the NJS1avg variable or its proxy along with
coefficient of variation(CV) of these estimates as
reflected in Table 2 below.
Table 2: Coefficient Estimates of Student’s STM
score in JSS1(NJS1avg) and School staff quality
Index(Nrsqindex), their standard errors,
coefficients of variation(CV) and model
deviance(D) for the
measurement error
unadjusted scenarios.
Table 2: Coefficients of Student STM
Data
Variable
Coefficient
estimate
Standard
error
1
NJS1avg **
Nrsqindex
NJS1avg*
0.273
-0.014
0.314
0.020
0.009
0.033
Coefficient of
variation
(CV)
0.073
-0.643
0.105
Nrsqindex
NJS1avg
Nrsqindex
NJS1avg
Nrsqindex
NJS1avg
Nrsqindex
-0.053
0.680
0.010
0.671
0.011
0.744
0.030
0.032
0.017
0.062
0.020
0.06 1
0.018
0.082
-0.604
0.025
6.20
0.030
5.54
0.024
2.73
2
3
4
5
Employing
Gibbs
Sampling
technique
implemented in MLWiN package 2.20 [10], we
adjust for the incidence of measurement errors to
obtain estimate results of the two predictor
variables under investigation as in Table 3 below.
Model
deviance
(D)
21967
15206
808
1188
3456
Table 3: Coefficient Estimates of Student’s STM
score in JSS1 (NJS1avg) and School
Staff quality Index (Nrsqindex), their standard
errors, coefficients of variation (CV) and model
deviance (D) for the measurement error adjusted
scenarios.
.
Table 3: Predictor Variables
Data
Variable
M.E
Variance
Coefficient
estimate
Standard
error
1
2
NJS1avg **
Nrsqindex
NJS1avg **
Nrsqindex
NJS1avg **
Nrsqindex
NJS1avg*
0.22
0.85
0.25
None
0.22
None
0.25
0.394
-0.101
0.416
-0.015
0.398
-0.015
0.468
0.029
0.074
0.029
0.009
0.028
0.009
0.10
Coefficient
of variation
(CV)
0.074
-0.732
0.070
-0.600
0.070
-0.600
0.218
2
Nrsqindex
NJS1avg*
0.85
0.25
-0.260
0.480
0.156
0.094
-0.600
0.196
Nrsqindex
NJS1avg
Nrsqindex
NJS1avg
Nrsqindex
NJS1avg
None
0.25
0.85
0.25
None
0.25
-0.039
0.871
0.035
0.860
0.043
0.828
0.027
0.029
0.070
0.019
0.050
0.021
-0.692
0.033
2.00
0.022
1.163
0.025
Nrsqindex
0.85
-0.007
0.041
-5.86
NJS1avg
Nrsqindex
NJS1avg
Nrsqindex
0.25
None
0.64
0.85
0.869
-0.003
0.893
-0.029
0.038
0.044
0.017
0.030
0.044
14.67
0.019
-1.03
NJS1avg
Nrsqindex
NJS1avg
Nrsqindex
NJS1avg
Nrsqindex
0.25
0.85
0.25
None
0.064
None
0.901
-0.025
0.909
0.005
0.899
0.002
0.016
0.028
0.014
0.012
0.014
0.013
0.018
-1.12
0.015
2.40
0.016
6.50
1
1
3
3
4
4
5
5
5
5
Following measurement error adjustments, Data
1, 3, 4 and 5 all reflected an average CV of the
coefficient estimate of NJS1avg to be equal to or
less than what obtained in the measurement error
unadjusted scenarios. In the case of the Nrsqindex
variable, the measurement error adjustment did not
seem as impressive as what obtains in the NJS1avg
Model
deviance (D)
20023
19904
20068
12732
12928
-4476
-4406
-9170
-9325
-16900
-17433
-18359
-17855
variable; Data 1 and 2 did not reveal a drop in the
numerical value of the CV of the coefficient
estimate of the Nrsqindex variable. Data 2, 3 and 5
however reflected numerical CV values of the
coefficient estimates of Nrsqindex
for the
measurement error adjusted cases to be, on
average, less than or equal to what obtained in the
.
measurement error unadjusted cases. We observe
that , apart from the near-absence between cluster
variations in so far as the Nrsqindex variable was
concerned, the variable also has a fixed coefficient.
It is discernable that, in general, measurement error
adjustments done gave rise to increase in numerical
size of perceived error-prone predictors, increased
standard error and reduced model deviance as
expected. It is also found that, in the measurement
error adjustments where measurement error
variance values were assumed( rather than
estimated) , coefficients tended to have been
inaccurately determined with exaggerated estimates
and lower standard errors. Ignoring the likelihood
of measurement errors in some predictors and
adjusting for error in some other predictors tend to
also yield much higher coefficient estimate values
with the overall model deviance not necessarily
being lower. Deviations from expected postmeasurement error adjustment effects are also
discernable for variables with low reliability (i.e
high measurement error variance); thus assuming a
low measurement error variance value (i.e. high
reliability) for a variable is likely to result in a
coefficient estimate value indicating a higher
predictive power than what obtains when we
assume a higher measurement error variance (i.e.
low reliability) for such a variable.
Regardless of some inadequacies arising from
supplementary data generation approaches that
gave rise to Data 4 and 5 above and hence seeming
unimpressive results in some measurement error
adjustments done, the hypothesis that using
estimated measurement error variance as input into
model estimation process, as done here using
bootstrapping and Gibbs sampling, is more
objective, more logical and realistic than using an
assumed value.
4.0 Conclusion
Although the incidence and effects of measurement
errors on the response and explanatory variables in fixed
effects models such as linear and generalized linear
models , has been well documented in the literature(see,
for example,[4],[9], [[2], [[8], [1], [12], [5]), studies on
the behaviour of biases associated with measurement
error in covariates or the response for mixed models
such as multilevel hierarchical linear models is, up to
date, not well known and can be complex [7]. One of the
daunting challenges that often confront researchers is
that of realistically estimating measurement error
variances and reliabilities of error-prone variables in a
multilevel model to enable realistic measurement error
adjustment. An iterative measurement error adjustment
technique entailing bootstrapping and Gibbs Sampling is
applied on an educational illustrative data (i.e Data 1-5)
to which levels two or four models are associated.
Employing the iterative measurement error
adjustment technique on the STM score per student in
JSS1 subjects (NJS1avg) variable generally indicated
numerical increase in the coefficient estimate, increased
standard error of the coefficient estimate, decreased
overall model deviance, decreased estimate of the
coefficient of variation(CV) of the coefficient estimate.
The near-absence between cluster variance coupled with
possible weaknesses in the supplementary data
generating simulation method employed in respect of
predictors with fixed coefficients(such as Nrsqindex)
and some data sets however revealed slightly differing
trends. It is opined in this paper that the bootstrapping
and Gibbs Sampling measurement error adjustment
approach for addressing incidence of measurement
errors in multilevel models is more efficacious in a
situation where the error-prone predictor variables under
consideration have random coefficients. It is a suggested
that a realistic appraisal of the effectiveness or otherwise
.
of a measurement error variance estimation and
measurement error adjustment approach should, apart
from examining the general expectations of increase in
numerical value of coefficient estimate, increased
standard error, reduced level 1 residual and reduced over
all model deviance, also take into cognizance the
coefficient of variation(CV) values of the coefficient
estimates associated with the perceived error-prone
predictors.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
Carroll, R. I., Ruppert, D., Stefanski, L.A. (1995) Measurement error in Nonlinear
Models. Chapman and Hall.
Degracie, J.S. and Fuller, W. A. (1972) Estimation of the slopes and analysis of variance
when the concomitant variable is measured with error. Journal of American Statistical
Association, 67, 930-937 .
Ecob, E. and Goldstein, H.(1983).Instrumental variable methods for the estimation of test
score reliability. Journal of Educational Statistics, Vol. 8, No. 3, 223-24.
Fuller, W.A.(1987). Measurement Error Models. Wiley, New York
Fuller, W. A. (2006). Measurement Error Models. New York, Wiley
Goldstein, H. (2003). Multilevel Statistical Models, 4rd edn. London; Edward Arnold: New
York, Wiley.
Goldstein,H.,Kounali,D,.andRobinson,A.(2008). Modeling measurement errors and
category misclassifications in multilevel models. Statistical Modeling, Vol. 8, No. 3, 243-261
Joreskog, K. G. (1970) A general method for analysis of covariance structures. Biometrika,
57, 239-251.
Plewis, I. (1985). Analyzing change: Measurement and explanation using longitudinal data
. NewYork, Wiley
Rasbash,J.,Browne,W.,Healy,M.,Cameron,B. and Charlton,C.(2010). The MLwiN
Command Interface version 2.20. Centre for Multilevel Modelling,University of Bristol,
Bristol.
Sargan,J.D.(1958). The estimation of economic relationships using instrumental variables.
Econometrica, 26,393-415.
Skrondal, A. and Rabe-Hesketh, S. (2004) Diet and heart disease: A covariate measurement
error model In Generalized latent variable modeling: Multilevel and structural equation
models,Chapman and Hall.
Woodhouse,G.(1998). Adjusment for Measurement Error in Multilevel Analysis.
London,Institute of Education,University of London.
.
Design and Implementation of an M/M/1 Queuing Model Algorithm and its Applicability
in Remote Medical Monitoring
Ifeoma Oji+ Osuagwu O.E. *
+Petroleum Training Institute, Effurum, Delta State
*Department of Information Mgt Technology, Federal University of Technology, Owerri
Abstract
Remote Medical Monitoring is a component of telemedicine capable of monitoring the vital signs of patients in
a remote location and sending the results directly to a monitoring station. Vital signs are collected by sensors
attached to the human body and sent automatically to the server in the hospital. This paper focuses on the
design and implementation of an M/M/1 queuing model capable of queuing the readings of the vital signs in the
server according to how they arrive on a First In First Out (FIFO) basis and sending them in turn to the
medical personnel when the need arises. The queuing model follows a Poisson distribution with parameter (β)t
and a probability function called the negative exponential distribution. The obtained output is based on a
simulation using the Queuing Model Simulator (QMS), simulation software which computes the mean, variance
and the total cost of running the queue.
Keywords: M/M/1 queuing model, FIFO, QMS, simulator, mean, variance, total cost
________________________________________________________________________________
1.0 Introduction
With the advancement of wireless
technologies, wireless sensor networks can
greatly expand our ability to monitor and track
conditions of patients in the healthcare area [8].
A medical monitor or physiological monitor or
display, is an electronic medical device that
measures a patient's vital signs and displays the
data so obtained, which may or may not be
transmitted on a monitoring network.
Physiological data are displayed continuously
on a CRT or LCD screen as data channels along
the time axis. They may be accompanied by
numerical readouts of computed parameters on
the original data, such as maximum, minimum
and average values, pulse and respiratory
frequencies, and so on [5].
In critical care units of hospitals, bedside units
allow continuous monitoring of a patient, with
medical staff being continuously informed of the
changes in the general condition of a patient [1].
Some monitors can even warn of pending fatal
cardiac conditions before visible signs are
noticeable to clinical staff, such as arterial
fibrillation or premature ventricular contraction
(PVC). Old analog patient monitors were based on
oscilloscopes, and had one channel only, usually
reserved for electrocardiographic monitoring
(ECG). So, medical monitors tended to be highly
specialized [8]. One monitor would track a patient's
.
blood pressure, while another would measure pulse
oximetry, another ECG [2]. Later analog models had
a second or third channel displayed in the same
screen, usually to monitor respiration movements
and blood pressure. These machines were widely
used and saved many lives, but they had several
restrictions, including sensitivity to electrical
interference, base level fluctuations, and absence of
numeric readouts and alarms. In addition, although
wireless monitoring telemetry was in principle
possible (the technology was developed by NASA
in the late 1950s for manned spaceflight. It was
expensive and cumbersome.
Typically a queuing model represents the
system’s physical configuration by specifying the
number and arrangement of the servers, which
provide service to the customers, and the stochastic
nature of the demands, by specifying the variability
in the arrival process and in the service process [9].
To achieve remote medical monitoring, patients’
data on vital signs are collected via sensors attached
to the patient’s body and sent automatically to the
server in the hospital. It is pertinent to note that
several patient’s information arrive to the server
and therefore it is very important to let these patient
data enter a queue from where they can be sent to
the different doctors assigned to do that. The queue
used here is the First In First Out (FIFO) queue.
2. Components of a basic Queuing System
Input Source
Calling
Population
The Queuing
Jobs
Queue
Arrival
Process
Service
Mechanism
Served
Jobs
Leave the
system
Queue
Discipline
Queue
Configuration
Service
Process
Fig. 2.1: Components of a queuing system
.
2.2 The Queuing Model for distribution of Patient data
Patient
IPD
A
Bluetooth
WBA
N
1
2
Route
r
WiF
i
Hub
PID
1
PID
2
PID
3
Doctor contacts patient through SMS for advice
Doctor
SP 1
Doctor receiving
patient’s data
Doctor
SP 2
Doctor
SP 3
Fig. 2.2: The Queuing Model Architecture
PID 1 Patient ID 1: This can be from Pid 1 to Pid n.
SP 1 Specialist 1: This can also be form Sp 1 to Sp n.
.
Patient
Router
IPDA
Hub
The Queuing
System (FIFO)
Doctor contacts patient through SMS
Doctor receiving
patient’s data
SP 1
SP 2
SP 3
Fig. 2.3: Block diagram of the queuing model
The M/M/1 queue consists of a server which
provides service for the packets of data from the
patients who arrive at the system and depart. It is a
single-server queuing system with exponential
interarrival times, exponential service times and
first-in-first-out queue discipline [4]. If a packet of
data from a patient arrives when the server is busy,
it joins the queue (the waiting line). There are two
.
AT = arrival time
DT = departure time
types of events: arrival events (A) and departure
events (D). The following quantities are used in
representing the model:
Poisson
Arrivals
Queue
Departure
Server
Fig 2.4: M/M/1 Queue
STA
INITIALIZE VARIABLES
YES
NO
AT<DT
C
A
ARRIVAL EVENT
DEPARTURE EVENT
D
B
NO
STOP
YES
COMPUTE MEAN VALUES OF OUTPUT
STOP
Fig 2.5: Flowchart of the queue simulation
.
Simulated time
Scheduled
departure time
First departure
T0
0
T1
Interarrival time
Next arrival
First arrival
Fig. 2.6: The first few events in simulation
In the queuing model, vital signs are collected by
the sensor on the patients’ body, sent through the
Bluetooth (this data is simulated in the IPDA) to the
IPDA which transmits this data by WiFi to the
router which further transmits the data wirelessly to
the hub which is in the server.
The hub acts as a data repository where these
data are stored and sent to the doctor when there is
an abnormal situation.
2.1.2 The Queuing Model
Queuing models can be represented using
Kendall’s notation.
A/B/S/K/N/D [3].
where
A is the interarrival time distribution
B is the service time distribution
S is the number of servers
K is the system capacity
N is the calling population
D is the service discipline assumed
•
The Arrival Rate
The data arrive as packets of data from different
patients wearing the sensors into the hub.
Let Ci be the interarrival time between the
arrivals of the (i – 1)th and the ith patients, the
mean(or expected) inter-arrival time is denoted by
E(C) and is called β; = 1/E(C) the arrival frequency.
•
Service Mechanism
This is specified by the number of servers
(denoted by S) each server having its own queue or
a common queue and the probability distribution of
the patient’s service time [7].
.
Let Si be the service time of the ith patient, the
mean service time of a customer is denoted by E(S)
=
Queue Discipline
Discipline of a queuing system means the rule
that a server uses to choose the next patient from
the queue (if any) when the server completes the
service of the current patient [6].
The queue discipline for this system is
Single Server- (FIFO) First In First Out i.e.
patient’s data are worked on according to when
they came to the queue.
•
Measures of Performance for the
Queuing System
Let
Di be the delay in queue of the ith patient
Wi be the waiting time in the system of the ith
patient
F(t) be the number of patients in queue at time t
G(t) be the number of patients in the system at time
t = F(t) + No of patients served at t.
Then the measures,
and
W=
are called the steady state average delay and the
steady state average waiting time in the system.
Also the measures,
F=
•
Single Channel Queue
[M/M/1] : {FCFS or FIFO} Queue System
the service rate of a server.
•
D=
time average number in the system.
and
G=
•
Arrival Time Distribution
This model assumes that the number of arrivals
occurring within a given interval of time t, follows
This
a poisson distribution with parameter
parameter
is the average number of arrivals in
time t which is also the variance of the distribution.
If n denotes the number of arrivals within a time
interval t, then the probability function p(n) is given
by
P(n) =
n = 0,1,2….
(1)
The arrival process is called poisson input
The probability of no(zero) arrival in the interval
[0,t] is,
Pr (zero arrival in [0,t]) =
= p(0)
Also
P(zero arrival in [0,t]) = P(next arrival occurs after
t)
= P(time between two successive arrivals exceeds t)
Therefore the probability density function of the
interarrival times is given by,
for t > 0
This is called the negative exponential distribution
with parameter
or simply exponential
distribution. The mean inter-arrival time and
standard deviation of this distribution are both
1/
where,
is the arrival time.
are called the steady state time average number in
queue and the steady state
.
Let the steady state probabilities be denoted by
Pn, n = 0,1,2,3,… where n refers to the number in
the system. Pn is the probability that there are n
units in the system. By considering a very small
interval of time h, the transition diagram for this
system can be seen as:
3.0Analysis of the Queuing System
The state of the queuing system can be
completely described by the number of units in the
system. Thus the state of the process can assume
values 0,1,2… (0 means none in the queue and the
service is idle).
βh
0
1
Ph
βh
βh
βh
βh
βh
n-1
2
Ph
n+
1
n
Ph
Ph
Fig. 4.18: The Transition Diagram
If h is sufficiently small, no more than one arrival
can occur and no more than one service completion
can occur in that time. Also the probability of
observing a service completion and an arrival time
.h2 which is very small (approximately
in h is
zero) and is neglected. Thus the following four
events are possible:
1. There are n units and 1 arrival occurs in h
2. There are n units and 1 service is completed
in h
3. There are n-1 units and 1 arrival occurs in h
4. There are n+1 units and 1 service is
completed in h
For n > 1, (because of steady state and condition)
Pr (being in state n and leaving it) = Pr(being in
other states and entering state n) = Pr(being in state
n-1 or n+1 and entering state n).
Thus
Pn
(2)
This is the steady state balance equation.
For n = 0, only events 1 and 4 are possible,
Po
1*
Therefore
P1 =
P0
Pn =
Pn-1
Pn =( )n P0
(3)
.
This can be determined by using the fact that the
sum of the steady state probabilities must be 1.
Therefore,
P0+P1+P2+…+Pn+Pn+1+…= 1
P0 + P0[
+ P0[
2
+…+P0[
n
+P0[
n+1
+ …= 1
P0[1+P + P2 + … + Pn + Pn+1 +…] = 1 P =
This is the sum of a geometric series. Therefore,
P0
= 1 as n
D=W- =
.
The traffic intensity P (sometimes called
occupancy) is defined as the average arrival rate
(lambda) divided by the average service rate (mu).
P is the probability that the server is busy.
P=
The mean number of customers in the system (N)
can be found using the following equation:
Since P< 1,
P0 = (1 – P) =
The term P =
is equal to the probability that the
service is busy, referred to as Pr (busy period).
4.0
Performance Measures
The average number of units in the system G can
be found from
G = sum of [n*Pn] for n = 1 to
G=
=
where P =
The average number in the queue is
F = (G – (1 – P0)
Sum of [(n-1)*Pn] for n = 1 to
T=
=
F=
The average waiting time in the system (time in the
system) can be obtained from
W=
You can see from the above equation that as p
approaches 1 number of customers would become
very large. This can be easily justified intuitively. p
will approach 1 when the average arrival rate starts
approaching the average service rate. In this
situation, the server would always be busy hence
leading to a queue build up (large N).
Lastly we obtain the total waiting time (including
the service time):
=
In a queuing system with the inter arrival time of
25 seconds and the service time of 10 seconds, the
parameters are calculated thus:
and
.
Table 1
E( C ) = 1/Ci
β
0.04
50
75
100
125
150
175
Ci
25
Si
E(S) = 1/Si
µ
P = β/ µ
N = P/1-P
T = 1/( µ - β )
10
0.1
0.4
0.666667
16.66667
0.02
20
0.05
0.4
0.666667
33.33333
0.013333
30
0.033333
0.4
0.666667
50
0.01
40
0.025
0.4
0.666667
66.66667
0.008
50
0.02
0.4
0.666667
83.33333
0.006667
60
0.016667
0.4
0.666667
100
0.014286
0.4
0.666667
116.6667
0.005714
70
200
0.005
80
0.0125
0.4
0.666667
133.3333
225
0.004444
90
0.011111
0.4
0.666667
150
250
0.004
100
0.01
0.4
0.666667
166.6667
This is illustrated in the line chart and column chart below:
.
Fig. 5.0
5.0 Results
The output was obtained from the simulation done using the QMS simulator.
Fig. 5.1: Results of a queuing model simulation
.
Fig. 5.2: Variance for the queue
Fig. 5.3: Average for the queue
.
Fig. 5.4: Max, Min for the queue
Fig. 5.5: Confidence Interval for the queue
6.0 Conclusion
A queuing model has been designed and simulated.
The result of the simulation has been shown. It has been
established that the packets of data from a patient’s body
arrives to the hospital’s server and enter the queue from
where they are kept according to the order in which they
arrive and any data with abnormal readings are sent to
the doctor for immediate medical intervention.
.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
Gao, T. (2005) “Vital Signs Monitoring and Patient Tracking Over a Wireless Network,” IEEEEMBS 27th Annual Int. Conference of the Eng. in Medicine and Biology.
Jurik, A.D.; Weaver, A.C. (2008). Remote health care monitoring devices Computer 41(4).
Kendall, D. (1953). "Stochastic Processes Occurring in the Theory of Queues and their Analysis
By the Method of the Imbedded Markov Chain". Annals of Mathematical Statistics 24 (3).
Lee, A., Miller D. (1966). "A Problem of Standards of Service ". Applied Queueing Theory. New
York: MacMillan.
Obrenovic, Z., Starcevic, D., Jovanov, E., & Radivojevic, V. (2002). An Agent Based
Framework Medical Devices. Autonomous Agents & Multi-Agent Systems, Bologna, Italy.
McGraw Hill Company.
Sen, R.(2010). Operations Research: Algorithms and Applications. Prentice-Hall.
Tijms, H.(2003), Algorithmic Analysis of Queues, A First Course in Stochastic Models, Wiley,
Chichester,.
Varshney U. (2008), “Improving Wireless Health Monitoring Using Incentive-Based Router
Cooperation,” Computer, 41(3).
Zhou, Y., Gans, N.(1999). A Single-Server Queue with Markov Modulated Service Times".
Financial Institutions Center, Wharton, UPenn. Retrieved from
http://fic.wharton.upenn.edu/fic/papers/99/p9940.html. Retrieved 2011-01-11.
.
Classification of Implemented Foreign Assisted Projects into Sustainable And Nonsustainable Groups: A Discriminant Analysis Approach
Iwuagwu Carmelius Chibuzo
E-mail cciwuagwu@yahoo.com Tel +2348033306938
Department of Planning, Research and Statistics. Ministry of Education,
Owerri, Nigeria.
Abstract
Four variables of factors affecting the sustainability of foreign assisted projects at the end of implementation
were extracted from literature review and informal interview of project management professionals. The
managers of these projects were asked to rank these variables according to their relevance and importance in
enhancing the sustainability and non sustainability of foreign assisted projects in Nigeria. Discriminant
function analysis was applied in carrying out detailed analysis of these factors. Two factors were found to be
the most discriminating factors among the four factors. The two factors are delivery of service or benefits and
long term institutional capacity. The study also further revealed that continued delivery of services or benefits
is the most discriminating factor.
Key Words: Economic transformation, discriminant analysis, foreign assisted projects, delivery
of service, long term institutional capacity
1.0
Introduction
Rapid economic transformation in developing
countries is often constrained by shortage of capital
resources due to poor savings, inadequate foreign
exchange earnings and low investment capacities.
There is therefore, the tendency for developing
countries to use more resources than those available
to them. It is this “resource gap” that has given rise
to the need for external development assistance
from developed world. Nigeria has been a major
beneficiary of development assistance in the form
of concessionary and non-concessionary loans,
outright grants and technical assistance. In Nigeria,
most of this assistance comes from UNDP and
UNICEF. These grants are used in financing and
implementing some development projects and they
are termed foreign assisted projects.
Many of these projects cannot be sustained at the
end of their successful implementation and handing
over to the beneficiaries. Many of these projects
survive for less than two years at the end of
implementation and stop functioning.[7]
Many reasons like bad implementation method,
frequent change in government, inadequate
funding, lack of beneficiaries support and other
environmental factors have been advanced for this
ugly trend[4],[10].
.
The study is concerned with building a model
which can be used to classify implemented foreign
assisted projects into one of two categories
sustainable and non-sustainable group. Although
the discriminant analysis which is used can be
generalized for classification into a number of
categories, the present study is limited to two
categories.
2.0
Applications
The discriminant analysis has been applied in a
variety of investigations. These applications vary
from parent selection in plant breeding to the
classification of application for bank loan into good
and bad creditors.[2][9][11][8][5] Discriminant
analysis used to discriminate between two
expenditures groups using the percentages of total
household expenditure going to five major budget
items as criterion variables.[14] As shown in the
study, giving a household’s percentage expenditure
on each of the five categories of commodities –
accommodation, food, transport, household goods
and clothing it is possible to use the household
discriminant score to determine the household’s
expenditure class – lower or middle.[14]
Many of the standard applications of the
techniques are found in the biological sciences, but
it is also potentially fruitful in the social sciences.
The technique was applied in an attempt to identify
underdeveloped countries with good development
potential. In their analysis, 73 underdeveloped
countries were classified into three groups
according to their past economic performance and a
linear discriminant function estimated from a
number of social, political and economic
variable.[1] Once such a function has been
estimated the values of these variables for a new
country can be fed into the discriminant function
and the country assigned to one of the three groups,
for development potential.[1]
The discriminant analysis was used to classify
household in Ile-ife into higher and “Lower”
income brackets. A linear compound of five
criterion variables namely, type of dwelling, rent
status, availability of possession of selected
household goods, size of household and highest
educational level was formed.[8] Education level
was found to have the highest weight in the
discriminant function. This study was found useful
where there are difficulties as directly obtaining
authentic information on household incomes.[8]
Discriminant analysis as a predictive tool for
corporate technical failure and bankruptcy, his
work provided answer to which ratios are important
in detecting corporate financial failure potentials.[3]
He utilizes a comprehensive list of financial ratios
in assessing a firms failure potential[3].
3.0. Sources of Data
The data were obtained from foreign assisted
projects UNDP and UNICEF located in six states
representing the six geopolitical zones of Nigeria.
The states are Imo, Rivers, Oyo, Borno, Kano and
Plateau. The set of data was extracted from
Questionnaire administered to the various projects
establishments responsible for the day to day
maintenance and sustainability of this foreign
assisted project when they are handed over to the
beneficiaries.
Seventy-one (71) implemented foreign assisted
projects of UNDP and UNICEF made up of 40
sustainable and 31 non-sustainable were used in the
analysis. Four criterion variables were used in
classifying implemented foreign assisted projects
.
into sustainable and Non-sustainable groups. These
criterion variables were:
U1
=
Delivery of service or
benefits
U2
=
Political support
U3
=
Long
term
institutional
capacity
U4
=
Maintenance of physical
infrastructures
Let Ci denote the covariance matrix for the same
sample calculated using the sample variance given
by
2
S 2j = ∑ ( Xij -X j ) / n − 1
n
In addition, the sample covariance between
variable j and k defined as
∑(x
n
Cjk =
ij
i =1
4.0. Discriminant Analysis
The problem that is addressed with discriminant
function analysis is how well it is possible to
separate two or more groups of individual given
measurements for these individuals on several
variables.[6] Two methods of discriminant function
analysis namely the Mahalanobis distance
(independent variable) and stepwise methods were
used.
4.1
Mahalanobis Distance Method
Let X 1 = (X1i , X 2i ...X pi ) denote the vector of
mean value for the sample from the its group
calculate using the
n
X j = ∑ Xij /n
j=1
− x j ) ( X ik − X k ) / n − 1
(4)
This being a measure of the extent to which the
two variables are linearly related.
The matrix of variance and covariance are given by
 C11 C12 ... C1p 


C =  C21 C22
C2p 
(5)
 C C ... C 
p2
pp 
 p1
Let C denote the pooled sample covariance matrix
determined using
m
C = ∑ ( n i -1) Ci
i=1
n
∑ ( n − 1)
i =1
(6)
i
Then the Mahalanobis distance from an observation
X’ = (X1, X2, …Xp) to the centre of group is
estimated as
(1)
D12 = ( X-X i )
X1 
 
• 
 
X = • 
• 
 
 X p 
or
−1
∑∑ ( X
p
Di2 =
(2)
(3)
i=1
C-1 (X-X i )
p
r-1 s −1
r
(7)
− X ri ) C rs ( X s − X si )
(8)
Where Crs is the element in the rth row and Sth
column of C-1. The observation X is allocated to the
group for which Di2 has the smallest value.
.
Specifically for this study, a total of three
canonical discriminant function were obtained. For
the sustainable and non-sustinable implemented
foreign assisted projects in Nigeria we have
Z = a1U1 + a2U2 + a3U3 + a4U4
(9)
And Z1 = a11U1 + a12U2 +a13U3+a14U4
(10)
Z2= a21U1+a22U2+a23U3+a24U4
(11)
Where Z denote the grouping that is for project
sustainability and project non sustainability and Z1
for project sustainability and Z2 for project nonsustainability.
The ai and aij – values are the canonical
coefficients/variables.
Therefore it is possible to determine several
linear combinations for separating group.[6]
Finding the coefficient of the canonical
discriminant functions turns out to be an eign value
problem. The within-sample matrix of sum of
squares and cross products is calculated using.
Wrc =
n
ni
j=1
i =1
∑ ∑( X
i jr
− X jr ) ( X ijc − X jc )
(12)
.
And Wrc is the element in the rth row and
cth column of W
The total sample matrix of sum of squares
and cross product T is calculated using
trc=
nj
∑ ∑( X
m
j =1
i =1
ijr
− X r ) ( X ijc − X c )
(13)
The in between group matrix is given by
B=T–W
(14)
Which can be determined.
The matrix W-1B is found. If the eigen value are
A1>A2>A3…>Ai the Ai is the ratio of the in
between group of sum of squares to the within
group of sum of squares for the ith linear
combination, Zi, while the element of the
corresponding eigenvector ai = (ai1, ai2 – aip) are the
coefficient of Zi.
5.2
Stepwise Method
In this method, variables are added to the
discriminant function one by one until it is found
that adding extra variable does not give significant
better discrimination. There are many different
criteria that can be used for deciding on which
variables to include in the analysis and which to
miss out.[13] The order in which the repressors are
introduced may be determined in several ways, two
of the commonest are:
(a)
The researcher may specify a prior the
order in which he wants the repressor to
be introduced
(b)
The researcher may want to let the
data determine the order.
For this study the WILK’s Lambda criterion was
used as the criterion for entering the equation.
The Wilk’s Lambda ( λ ) is defined as
=
se
ST
(15)
where the matrix Se is the error of squares and
cross product matrix ST is the within sum of square
and cross product. (SSCP) matrix for the r samples.
.
Similarly the matrix ST is the total SSCP matrix.
This is the matrix of sets of squares and cross
products of our entire combined samples regardless
of which population give rise to the sample items.
As in ANOVA we have the relation
ST = S λ + Se
(16)
Where S λ is the among SSCP matrix. The SSCP
matrix is defined
 ∑ W 12 ∑ W 1 W 2 ... ∑ W 1 W k 


S =  ∑ W 2 W 1 ∑ W 22 . . . ∑ W 2 W k 

2 

 ∑ W kW 1 ∑ W k W 2 ... ∑ W k 
.(17)
Major statistical packages generally have a
discriminant function for the application of the
methods as described in Equation 1-17. The data
were analyzed using the SPSS Program
Discriminant Version 10. Two methods of selecting
discriminating variables are available in this
software packages namely, the independent
(Mahalanobis) method and stepwise procedures.
However analysis here was carried out using step
wise procedures.
5.2.1 Evaluation of The Performance Of The
Model
The evaluation of the performance of the
classification of the discriminant model was based
on some statistical criteria, validation and out of
sample results. The statistical criteria employed
included F-value, Eigen value, Wilk’s Lambda,
Chi-square and canonical correlation.[13] The
validation (that is in sample or resubstitution) test
utilized the same set of sample observation while
the out of sample was based on completely
independent set of sample from the ones used in the
model estimation.[12]
In each case, we obtained the error rates
associated with the model in addition to the overall
error rates and overall correct classification rates.
5.3
Cutting Scores
The cutting score is zero. Discriminant scores
greater than zero (ie the scores) indicated a
predicated membership in the sustainable group.
The dependent variable which was continuous
scaled took the value zero and 1 for the nonsustainable group and sustainable group
respectively.
5.3.1 Relative Discriminatory Power Of The
Variables
The magnitude of the discriminant coefficient in
the model reveals to some extent the importance of
the discriminatory variable. However a major
objective procedure of evaluating the contribution
of each discriminating variable to the model is
based on the relative discriminating power of the
coefficient in the canonical discriminant
function.[9] The measure of the relative
discriminatory power of the variable is given by bi
( X i1- X io ), the scalar vector biσ i is used to
measure how the variables are correlated.
Here bi = the discriminant function coefficient for
the ith variable.
δi
=
The square root of appropriate value
in the variance – covariance matrix (standard
deviation)
X i1
=
ith variables mean for the
Successful project
X i0
= ith variables mean for t failed project.
.
7.0 Model Estimation And Interpretation
The data described above was used to estimate
the canonical discriminant function. Thus we used
(i)
Canonical Discriminant function (W)
W = -11.259+0.449U1+0.278U3
stepwise procedure and the result of
discriminant analysis is presented in table 1.
the
(18)
Table 1: Sustainable and Non-sustinable Foreign Assisted Projects
Eigenvalue
Wilkslambda Chi-square
Significance F-value
Canonical
correlation
9.688
0.094
161.100
0.000
7.520
0.952
(ii)
(iii)
Classification Function coefficients
Variable
Sustainable Projects
U1
5.348
U2
5.299
Constant
-102.319
Group Centriods
W
0.00
1.00
Based on the summary statistics presented in
table 1,we found that only two variables out of four
variables
considered
were
adequate
for
discriminating implemented foreign assisted
projects into sustainable and non sustainable
categories. The ratio in order of importance based
on the magnitude of their coefficients was (a) U1
(b) U3. However a more objective procedure of
Non-sustainable projects
2.571
3.581
-35.091
Function
-3.485
2.701
evaluating the contributions of each discriminatory
variable to the model is based on the relative
discriminatory power of the coefficients in the
canonical discriminant functions. The relative
discriminatory power of the variables of this model
is shown in table 2.
.
Variable
U1
U3
Total
Table 2: Relative discriminatory power of the variables equation (18)
W = -11.259+0.449U1+0.278U3
σi
ai
ai σ i
U i1
U i0
ai (Ui1 − U i 0 )
0.449
5.244
19.625
9.774
2.354
4.423
0.278
3.634
18.550
12.194
1.010
1.767
6.190
Table 2 showed that variable U1 explained about
71.45% of the average discriminant score
separation between sustinable and non-sustainable
implemented foreign assisted projects while U3
contributed 28.55% in explaining the average
discriminant score separation between foreign
assisted project sustainability categories. The scalar
vector showed that the variables used in
constructing the discriminant model exhibit little or
no correlation among them.
7.0
Evaluation
of
The
Classification
Ability of The Discriminant Model
In table 1, we present the discriminant model
denoted as equation (18) with the associated
statistics. The eigen value for the model is 9.688
while the Wilk’’s Lambda was 0.94 which is little
low and the canonical correlation is 0.952. In order
to test the statistical significance of the model (DF),
the Wilk’s Lambda was converted into chi-square
distribution and the model was found to be
significant at 100% level.
%
71.45%
28.55%
100%
The canonical correlation value (CCV) of 0.952
implied a very high degree of association between
the discriminant function and the discriminating
variables U1 and U3.
The results of the validation test based on the
original samples for the discriminant model denoted
as W are presented in table 3. This table contains
actual and predicted discriminant scores sustainable
and non sustainable groups of implemented foreign
assisted projects. Based on table 3, only one
implemented foreign assisted projects was
misclassified in the sustainable group. The project
is number 12.
The out of sample result are presented in table 4.
Based on table 4, only three (3) out of eleven (11)
implemented foreign assisted projects were
wrongly negatively classified as non sustainable
projects and one (1) out of seven (7) nonsustainable foreign assisted projects was also
wrongly positively classified as sustainable project.
.
Table 3: Classification of Implemented Foreign Assisted Projects based on validation sample (41
sustainable vs 30 non-sustainable)
Project No
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
Group
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
Discriminant scores
2.5457
3.3788
3.8922
1.3703
4.7253
1.0926
2.0968
2.2680
2.9945
2.5457
2.5457
-0.2539
1.8191
0.5371
4.1699
0.6437
3.1656
4.1699
2.7168
2.2680
3.6145
1.8191
2.8234
2.9945
2.8234
3.1656
3.8922
2.2680
3.4433
2.7168
3.4433
3.8922
3.3732
2.2680
2.7168
2.2680
3.4433
2.7168
1.9902
-0.0914
.
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
3.3958
-2.2625
-3.2247
-3.6735
-3.7801
2.6468
-4.5066
-3.3313
-3.3203
-3.4127
-3.8867
-4.3355
-3.1601
-3.7801
-1.9848
-3.7156
-3.9933
-3.5445
-3.9933
-3.7156
-3.9933
-3.2667
-3.4379
-4.8910
-3.9933
-4.4421
-4.4421
-3.9933
-4.4421
-1.6004
-3.7156
.
Table 4: Classification of Foreign Assisted Projects Based on out of sample Data (11 Sustainable Vs 7
unsustainable Projects)
Project No
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
Group
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
Discriminant Scores
2.383
3.623
-.0632**
2.832
3.623
-1.359**
2.447
1.827
-0.354**
0.202
2.554
3.901
-2.086
-2.237
0.651*
-3.433
-3.540
-4.438
Source: Using the Developed model
W = -11.259+0.449U1+0.278U3
**………………wrong classification
In table 5 we present various error and
classification efficiency rates associated with the
discriminant model equation (18). the validation
sample method was extremely accurate in
classifying about 98.59% of the total sample
correctly. type 1 error proved to be only about 2.5%
while type ii error was about 0%.
The predictive ability of the model based on the
out of sample data showed that type i error
associated with the model was 25.99%. This
implied that about 25% of the sustainable
implemented foreign assisted projects were
wrongly classified as sustainable projects. however
the overall classification efficiency of the model
based on the out of sample data was high since
77.78% of the implemented foreign assisted
projects were correctly classified while only about
22.22% represented the overall error rate.
The high overall classification efficiency rate of
98.59% and 77.78% for validation and out of
sample procedures suggested that the model may be
useful as early warning device for predicting and
classifying implemented foreign assisted project
into
different
risk
categories
. Table 5: Classification Result for the Discriminant Model
Count
Out
sample
count
(b)
Validation
count
Implemented foreign assisted projects
of
Sustainable
(1.00)
Non-sustainable (0.00)
1.00
0.00
total
Sustainable
(1.00)
Non sustainable (0.00)
Predicted group members
1.00
9
1
75.00%
83.33%
66.78%
39
0
97.50%
100%
98.59%
0.00
3
5
25.55%
16.67%
22.22%
1
31
2.50%
0%
1.43%
Total
12
6
100%
100%
100%
30
31
100%
100%
100%
1
.
8.0
Conclusion and Recommendation
In this study, the discriminant function analysis
was used to classify implemented foreign assisted
projects in Nigeria into “Sustainable” and nonsustainable groups. A linear compound of two
variables namely, Delivery of services or benefits
and long term institutional capacity was formed for
the sustainable and non-sustainable group. They
were found to be the most important factors that
8.2
Recommendations
Based on the
above, the following
recommendations are offered on the basis of the
research funding.
1.
For continued delivery of service which is
the most discriminating factor for foreign assisted
project sustainability government should provide
enough budgetary allocation for the maintenance of
facilities that ensured continued delivery of service.
Part of this budget should be made available to the
benefiting communities to enable them manage
these projects located in their place effectively. In
the case of water projects, the community leaders
should be allowed to sell the water at a reduced rate
so as to have money to effect necessary repairs
without waiting for the government.
However, this should be monitored by the
government to avoid abuse by these community
leaders.
discriminate between the sustainable and nonsustainable group. Delivery of services or benefits
was found to have the highest weight in the
discriminate function. All the coefficients of the
variables have the expected sign and the overall
discrimininating abilities of the function was found
to be quite high as indicated by the various tests of
the performance of the model.
2.
Project sustainability will frequently require
an active involvement of local and community
organization at all the stages of project planning,
implementation and operations. The results of the
virtual exclusion of beneficiaries often become
apparent during the operational phase when
beneficiaries only prove unwilling to pay for
services or when they refuse to cooperate in project
maintenance. Hence the government should
promote beneficiaries participation which will have
the positive result of ensuring project sustainability.
3.
The role of donor agencies is crucial to
project sustainability. They should established long
term institutional capacity that ensure sustenance
when they handover the project. Long term
institutional capacity was found to be a very
important discriminating factor of sustainability in
this research. They should also offer advisory role
from time to time to ensure project sustainability
References
[1]
Adelman & L. T Morris (1968) “Performance Criteria for Evaluating
.
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
Economi9c Development Potentials – An Operational Approach” Quart J. Economics, 82 p.
268-280.
Aja Nwachukwu (2001) “A Discriminant Analysis of Commercial Banks
Failure: The Case Study of Nigeria” Journal of Business & Finance – Imo State Vol. No. 4
2001 pp 26-35.
Altman – E. (1968) “Financial Ratios Discriminant Analysis and the
Prediction of Corporate Bankruptcy” The Journal of Finance Vol. 23, September.
Cleland D.I. and Kerzner, H. (1985) A Project Management
Dictionary of Terms, Van Nostrand Reinhold, New York.
Deakin E. (1972): “A Discriminant Analysis of Predictor of Corporate
Failure” Journals of Accounting Research Vol. 10 Spring.
Hope Keith (1968) Method of Multivariate Analysis, University of London.
Iwuagwu C.C. (1994): “The Evaluation of the Effect of the Ten Critical
Success Factors on the Performance of ADP Projects”, Owerri, Unpublished M.Sc. Thesis,
FUTO Nigeria.
Iyaniwuru J.O. (1984) “Discriminant Analysis as a Technique for a
Dichotomous Classification of Urban Income Earners in Ile Ife Nigeria” Journal of the
Nigerian Statistical Association Vol. 2 No.1, 1984 Pp. 17-26
Joy M.O. and Tollefson J.O. (1997) “On the Financial Application of
Discriminant Analysis” Journal of Financial & Quantitative Analysis (December). Pp 723739.
Kayoed, M.O. (1978) The Art of Project Evaluation Ibadan University
Press, Ibadan.
Keleck, W.R. (1980) Discriminant Analysis Services Quantitative
Application in the Social Sciences. No 19 Suga University Paper.
Laucherbanch P.A. (1967) “An Almost Unbiased Method of Obtaining
Confidence intervals for the Probability of Misclassification in Discrimination Analysis”
Biometrics (December) pp. 639-645.
Lewis, Beek (1980): Applied Regression: An Introduction to
Quantitative Application in the Social Sciences 22 Beverly Calif; Saga.
Olayemi, J.K. & Olayide, S.O. (1977) “Expenditure Pattern in Selected
Area of Western Nigeria: A Discriminant Analysis” Journals of Statistical Research Vol. 11
1977.
.
A Study on the Evaluation of Industrial Solid Waste Management Approaches in Some
Industries in Aba, South Eastern Nigeria
Ajero, C.M.U and Chigbo,U.N
Department of Environmental Biology Imo State University, Owerri Imo State Nigeria.
medoayero@yahoo.com
Abstract
Industrial solid waste is a serious health concern in Aba, South East Nigeria. This study was undertaken to
assess the approaches of some industries toward some aspects of waste management in Aba. Interviews,
observation and questionnaires administered to industry executives and waste managers were used to generate
data for the study. The result from the field research showed that majority(47.39%) of the solid wastes stream
were compostable while far less (4.69-9.90%) recyclable waste were generated. The result revealed that open
ground was the most prominent storage facility used and only 42.50% of the industries undertake some form of
treatment of waste before storage. The industries studied employ more than one method of waste disposal and
it was mostly carried out with open trunk and wheel barrows. Waste contractors were engaged by 52.50% of
the industries and wastes were equally deposited in dump sites (45.00%), rivers (12.50%), composited
(20.00%)or incinerated(25.00%). There was inadequate provision of protective measure in most of the
industry. The respondents acknowledged enjoying medical check which was however not regular and in some
cases in poorly equipped health centers and drug stores. The awareness of respondents on the consequences of
improper waste management was high (75.50%) while the level of attendance of health and safety treatment
training was average (46.25%), however, none of the industries had health policy plan. It is recommended for
the improvement of the function of the waste management regulatory body, Abia State Environmental
Protection Agency(ASEPA) and commitment of the industries and the general public toward industrial solid
waste management.
Key words: Solid waste, Industries, Management processes, Aba, Nigeria
_____________________________________________________________________________
1.0 Introduction
In industrial production, many materials are
employed to manufacture products and because of
the scale of the production, a lot of waste is
generally produced. The component categories
usually include: compostable (includes food, yard,
and wood wastes); Paper; Plastic; Glass; Metal; and
other (includes ceramics, textiles, leather, rubber,
bones, inerts, ashes, coconut husks, bulky wastes,
household goods).[1][2][3]
.
The improper management of solid waste poses
health hazards to the residents causing diseases
such as bronchitis and cancer.[4] High level of
industrial discharge has effect of upsetting the
ecological balance of nature. The microbial
degradation of waste in water bodies lead to
depletion of oxygen and fishes and other aquatic
organisms which require oxygen for survival are
thus affected.[5] In the extreme cases there is
dislocation of socio-economic system of an area.
As a result of the size of the problem, industrial
companies employ waste managers to focus solely
on the issue of proper and effective disposal of
waste. Industrial waste management involves
collection, transport, processing or disposal,
management and monitoring of industrial waste
material. Management of non –hazardous,
residential and institutional waste in Nigeria is
usually the responsibility of local/state government
authorities while management of hazardous
commercial and industrial waste is usually the
responsibility of the generator.[ 6][7,][8]
The problems relating to management of
industrial solid waste is associated with lack of
infrastructural facilities and negligence of industries
and environmental regulatory body to take proper
control..[9] Most major cities in Africa have an
established municipal waste collection system.
Collection is carried out by human- and animaldrawn carts (wheelbarrows, pushcarts), open-back
trucks, compactor trucks, and trailers. Collection
rates across the continent range from 20 to 80%.
Common feature of the municipalities is that they
are ineffective, underequipped and poorly
maintained (often vehicle immobilization rates
reach as high as 70%), inadequately funded and
poorly staffed.[3] The large industries (industries
with huge infrastructure, high man-power
requirement and influx of capital asset) located in
identified industrial areas respond to industrial
compulsion as imposed by the pollution control
laws by having some arrangement to dispose their
solid waste. However, the problems persist with
small industries. They find it easy to dispose waste
here and there, thereby mixing industrial,
residential and commercial waste and making it
difficult for local bodies to collect such waste
though it is not their responsibility.[11] The situation
is disturbing since it is estimated that small scale
units put together generate as much waste as the
large industries [9] overwhelming majority of
landfills in Africa are open dumps..[3] These
facilities are generally located at the perimeter of
major urban centers in open lots, wetland areas, or
next to surface water sources. Though many
municipalities have statutory requirements for the
construction and maintenance of landfills these are
generally not enforced. In most instances the
landfills are owned and operated by the same public
agency that is charged with enforcing the standards.
Often a lack of financial and human resources,
coupled with absent enabling policies, limit the
extent to which landfills can be built, operated, and
maintained at minimum standards for sanitary
practice.
There is no significant waste recovery and reuse
activities in Nigerian cities. In most cases,
scavenging plays an important role on the economic
survival of a number of industries (e.g., steel, pulp
and paper).[2] Waste pickers work on dumps and
even landfills, while some build squatter colonies
on the edges of dumps, sometimes with disastrous
consequences. Waste pickers are involved in a
small-scale recovery and reuse operation.
.
The assessment of industrial waste management
problems greatly varies depending on the nature of
industry, their location and mode of disposal of
waste.[12] Sound waste management cycle helps in
reducing the adverse impacts on the human health
and environment, while enhancing the lifestyle and
developing the economic state of the country. In
order to offer an appropriate solution for better
management of industrial solid waste in industrial
town like Aba in South- East of Nigeria, assessment
of the approaches of industrial waste management
is essential. This study aims to evaluate some
industrial waste management approaches of some
industries in Aba.
2.0 Study Area
The study was carried out in Aba in Abia State
of Nigeria. Aba is located at longitude 7°19¹E and
latitude 5°10¹N. It has a population of 839,000 and
accounting for four of the seventeen Local
Government Areas in Abia State.[13]
Aba known as ‘Japan of Africa’ is the
commercial and industrial center South East of
Nigeria, situated at the bank of river Aba. There are
many large and scale industries and four major
markets.
The solid waste management is carried out by
private
establishments
and
Abia
State
Environmental Agency. The landfill is situated at
the outskirts of the town; however there are
numerous dump sites. Most roads are filled with
refuse which leads to flooding especially during the
rainy seasons. The waste provides breeding ground
for vectors of human diseases and source of
unpleasant odour.
There are many private and government owned
hospital and healthy centers. There is very
ineffective public pipe borne water supply system.
3.0 Methodology
The researcher adopted several instruments in
data collection for this study in 2009. Planned
questionnaires (including questions on nature of
waste, storage of waste, collection, and method of
disposal, treatment and waste manager’s welfare)
were administered to waste managers, personnel
and managers of industries. Furthermore additional
data and information were collection from direct
observation, interview with responsible persons.
Secondary data were obtained from literature on the
subject. A total of 192 respondents were sampled
from twenty five small scale and fifteen large scale
industries
made
up
of
8(20,00%)
extracting,9(22.50%)
hospitality,
15(37,50%)
manufacturing 4(10.00%) construction and 4
processing industries. Ethical considerations such
as informed consent and confidentiality of personal
information in the interview were observed.
The percentage volume of the waste generated
by the companies is shown in Table
1.
.
Table 1: Types of the solid waste generated
Type of waste
Plastics
Papers
Glass
Aluminum scraps
Metal scraps
Compostable(eg.food and wood)
Water sachets and cellophane packages
Wastes that can be readily composited had the
highest volume of 47.39% while 4.69% was
recorded for glass. Other percentage volume
encountered were metal scraps 10.41%,paper and
Table 2:
Types of storage
Small scale
Industry
N=25 (%)
Uncovered drum
6(24.00)
Open ground
14(56.00)
Concrete bays
0(0.0)
Pilled against a wall 2(8.00)
Covered bin
3(12.00)
Volume(%)
12(6.25)
19(9.90)
9(4.69 )
19(9.90)
20(10.41)
91(47.39)
22(11.45 )
aluminum 9.90%,cellophone containers 11.45%
and plastic 6.25%. The response on the storage
facilities used by the industries is shown in Table 2.
Types of storage
Large scale
Total
industry
N=40
N=15 (%)
(%)
3(20.00)
9(22.50)
8(53.33)
22(55.00)
0(0.00)
0(0.00)
1(6.66)
3(7.50)
3(20.00)
6(15.00)
Open ground was the most prominent(55.00%)
and pilling against a wall accounted for 7.50% and
storage facility used by the industries followed by
15.00% of the storage facilities used.
uncovered drums(22.50%).The result revealed that
The result of interview and questionnaires on the
higher percentage of small scale industries
preliminary treatment of waste before disposal by
employed both facilities. None of the industries
the industries is illustrated in Table 3.
stored their refuse in concrete bays. Covered bin
Table 3: Preliminary treatment of waste before disposal
Questions
Any waste treatment
17(42.50)
Types of treatment
Segregation
Recycle/Reuse
Types of industry
Small scale(%)
Large scale(%)
No
Yes
No
Yes
18(72.00)
7(28.00)
5(33.33)
N=7
7(100)
2(28.57)
N=10
10(100)
3(30.00)
Total
(%)
No
10(66.66)
N=17
17(100)
5(29.410
Yes
23(57.50)
.
The result revealed that 42.50% of the industries
undertake some form of treatment with 28.00% of
large and 66.66% of small scale industries
involved. All the industries that reported treatment
of waste carry out segregation while only 29.41%
recycle or reuse their waste. The methods of
disposal of waste by the industries sampled are
displaced in Table 4.
Table 4: Method of disposal of waste
Disposal method
Waste collection contractor
Deposited at waste dump
Solid to other industries
Deposited in the river
Compositing
Incinerating
Types of industry
Small scale
Large scale
Total
N=25 (%)
N=15 (%)
N=40(%)
11(44.00)
10(66.00)
21(52.50)
8(32.00)
5(33.33)
18(45.00)
2(8.00)
6(40.00)
8(20.00)
5(20.00)
0(0.00)
5(12.50)
5(20.00)
3(20.00)
8(20.00)
8(32.00)
2(13.33)
10(25.00)
The industries studied employ more than one
method for waste disposal(Table 4). Twenty one
(52.50%) of the industries(10 large scale(66.00%)
and 11(44.00%) small scale) employ contractors for
waste disposal. Eight(32.50%) industries deposite
their waste at the municipal dump while 20.00% of
the industries either sale or composite their
industrial waste. Only 5(12.50%) small industries
deposite their waste in river while 10(25.00%)
carry out on-site incineration in make-shift
incinerators
Table 5: Preventive measures available to waste managers
Measures
Small scale
N=82(%)
Handkerchief
Respirator
Nose guard
Overall clothing
Jungle boot
Hand gloves
80(72.72)
0(0.00)
30(27.27)
12(10.90)
17(15.45)
23(20.90)
Number of respondents
Total
N=110(%)
Large scale N=192(%)
17(20.73)
10(12.19)
65(79.26)
55(67.07)
63(76.82)
74(91.24)
97(50.52)
10(5.20)
95(49.47)
67(34.89)
80(41.66)
97(50.52)
.
Table 5 shows the preventive measures available
to waste managers. From the result, 97(50.52%)of
the respondents acknowledge the use of hand
kerchief and handgloves.Jungle boot and nose
guard were provided to 41.66% and 49.47% of the
respondents respectively, while only 34.89% of the
respondents had overall clothing. Incidentally, only
10(5.20%) of the respondents from large scale
industries acknowledge the use of respirator.
Generally, large scale industries performed better in
the provision of preventive measures except in the
provision of unconventional handkerchief (72.72%)
by small scale industries. The respondents’
response on prophylative measures available in the
industries in shown in Table 6
Table 6: Prophylative measures available to waste managers
Variable
Number of respondents
Small scale
Large scale
N=110
N=82
O(0.00)
0(0.00)
0(0.00)
17(20.73)
14(12.72)
35(42.68)
25(22.72)
21(25.60)
41(37.27)
6(7.31)
30(27.27)
3(3.65)
Regular medical checkup
Once a week
Once a month
Once a quarter
Occasionally
Not at all
Total
0(0.00)
17(8.85)
49(25.52)
46(23.95)
47(24.47)
33(17.18)
Percentage in parenthesis
None of the waste managers had regular medical
checkup while 17(8.85%) from large scale
industries had weekly check up(Table 6). Forty
six(23.95%) and 49(25.52%) had quarterly and
monthly medical
check up respectively.
Respondents who reported occasionally check were
47(24.47%) of which respondents from small scale
industries accounted for
87.23%(41/47).Unfortunately,33(17.18%) of the
respondents of which majority, 90.90%(30/33) are
from small scale industries have not had any formal
medical check up. The table 7 refers to the response
of respondents on the availability of referral for
medical check up to waste managers.
Table 7: Response on referral available by the industries to the waste managers
Variable
Company clinic
General hospital
Health center
Near by drug store
Not available
Small scale
N=80
7(8.75)
18(22.50)
8(10.00)
6(7.50)
41(51.25)
Frequency(%)
Large scale
N=79
53(67.08)
18(22.78)
8(10.12)
0(0.00)
0(0.00)
Total
N=159
60(37.73)
36(22.64)
16(10.06)
6(3.77)
41(25.78)
.
Sixty (37.73%) of the respondents were referred to
company clinic of which 89.33%(53/60) are
workers of large scale industries. The result showed
that 22.64% respondents attended government
general hospital while 10.06% attended of the
health centers .Only 6(7.50%) of respondents and
from small scale industries workers were referred
to drug stores .Similarly, 41(25.78%) who reported
no referral to health facility were worker from small
scale industries. Table 8: Response on the
awareness of health consequences of industrial
solid waste and attendance of waste and attendance
of health/safety training on waste management
Table 8
Variable
Awareness of health
Consequences of waste
Attendance of health
And safety training
Health policy
Number of respondents from the industries
Small scale N=110
Large scale N=82
71(64.54)
73(89.02)
Total
N=192
144(75.00)
30(27.27)
59(71.95)
89(46.35)
0(0.00)
0(0.00)
0(0.00)
Percent in parenthesis
The result illustrated in table 8 shows that 75% of
the respondents were aware of the health impact of
improper waste management, however only 46.35%
of the respondents had attended health and safety
training on waste management. Unfortunately, only
27.27% of worker from small scale industries had
opportunity for the training.
Discussion
The industrialization of Aba has assisted in
building self reliant population and also in uplifting
of Nigerian economy. However, the huge waste
generated has caused serious problems relating to
environmental pollution. The problems relating to
the disposal of industrial solid waste are associated
with lack of infrastructural facilities, negligence of
industries and lack of commitment to take proper
safeguards.
The Abia State Environmental Protection
Agency has not been able to enforce the legal
provision and make industries legally responsible
for safety of all concerned. The component
category of the waste is semilar to other reports
from several authors in different cities.[1][10][11][14][15]
The result shows that compostables recorded the
highest volume of 45.75%. The high organic
content suggests possible value as composting
material. Composting is mainly practiced by the
hotels and eateries. This process converts waste to
manue for agricultural purposes. However, the
benefits are mostly not achieved as segregation is
mostly not practiced to remove the non-degradable
materials thereby producing low quality compose
.
.As a result, the viability of this scheme is
hampered by poor demand of the end markets for
their products. The waste stream indicates limited
potential commercial value for the recovery of
metals, glass, and plastic. This limitation does not
deter the uncoordinated and unhygienic scavenging
of waste in Aba.
The analysis of the type of storage for industrial
waste showed that little attention was paid to proper
storage as the wastes were mostly exposed in
uncovered container. The result in the study is
similar
to
the
observations
of
other
authors.[6][16][17][18][19] The hazardous and nonhazardous waste are mixed and are expected to
produce health problems among the workers and
handlers of waste and the general population.[4]. The
waste storage sites constitute foci for vectors of
diseases, source of our pollution and possibly
avenue for poisonous leacheates to contaminate
underground water.[21][22]. Equally, the unrestricted
access to dumpsites by unauthorized persons pose
health risk. [23][24] Unfortunately, the hygienic
covered bins were used by only 15% of the
Industries. This trend was observed in the large and
small scale industries revealing an institutionalized
disregard for the regulation on the use of covered
containers for waste storage.
Preliminary treatment of waste is necessary in
reducing the volume of hazardous waste and the
key to achieving sound industrial waste
management thereby encouraging source recovery,
possible reuse and health risk reduction. [27] The
57.5% prevalence recorded in this study among
industries without any form of pretreatment is
noteworthy, with 66.6% of them being large scale
as against only 16.0% small scale industries. It is
estimated that scale units put together generate as
much waste as large units [9] as most of the small
scale industries do not reduce their waste thereby
contributing enormously to the waste problem.
Investigations revealed that the recycling plants
engaged by 5% of the industries were purely for
paper recycling only. This limitation means that the
small recyclable waste is not recovered and the
concept of waste to health is completely neglected.
In Aba, the approach is to dispose waste as cheaply
as possible, without much concern as to what
happens once the waste leaves the production
premises. The industries surveyed in this study
employ more than one means for disposal of waste.
The waste collection contractors used by 5% of the
industries are in most cases engaged with disregard
to their capability. The wastes are manually
handled. There are few mechanical aids for waste
management. Stored waste is shoveled by hand into
open trucks or lorries encouraging fly-tipping and
often waste are taken to disposal sites impropriate
for the type of waste concerned. These unhygienic
practices are common in other cities in Nigeria.[23]
[6][11][8]
The reasons attributed to these include
financial limitations, corruption and illiteracy.
The result revealed that in 32.5% of the
industries, industrial solid waste are intermingled
with domestic waste making it difficult for waste
disposal agency to manage. The health implication
of this practice is magnified by the fact that nonhazardous waste assumes toxic nature once mixed
with hazardous materials. [23][24]
Five (20%) of small scale industries
acknowledged deposition of waste in the river side
thereby causing pollution and ecological
disruption.[5][25]
Fortunately, the large scale
industries which naturally produce most of the
hazardous waste do not deposite their solid waste in
.
the river. The small scale industries may have been
lacking in action to dispose off it’s waste and
upload it’s statutory responsibility due to lack of
education, awareness and trained personnel to
manage the waste in the industries, as well as
paucity of fund available to create a proper waste
management system.
It is a common knowledge that in Aba, small
scale industries do not seek the consent of the
regulatory body, Abia State Environmental
Protection Agency on waste disposal and equally
their un-wholesome activities are ignored to the
detriment of the inhabitants. It was clear from the
survey and interviews that it is likely that the
volume of solid waste will steadily increase as the
owners are not accountable to higher authorities nor
are the workers adequately informed. The results of
this study are similar to those of other studies
conducted in other countries.[26]. The government
should monitor these industries so that they can be
responsible with their waste management.
In the ten industries that carry out incineration,
the process is carried out in make-shift on site
incinerators operating at temperature below 800oC,
consequently inhabitants that live near the
industries are exposed to dioxin, and furan or other
toxic pollutants. [27]. The result further revealed
that the industrial waste mangers use several
protective measures. The large scale industries
faired better with 10 (12.1%) of the respondents
acknowledging to have used recommended air
respirator. However, the use of handkerchief in the
light of availability of hygienic and cheap nose
guard is embarrassing. The companies’ executives
need to be educated on the benefits of adequate
protective measures for their workers. The
improvement of the workers welfare will have
direct and positive relationship with their
productivity.
Medically, it is advised that waste mangers or
handlers subject themselves to prophylactic
measures to amenoriate unnecessary health
problems [28]. Even though most of the respondents
in this study acknowledged undertaking some form
of prophylactic measures, none had regular medical
check-up and as much as 17.1% mostly from small
scale industries have not had any form of medical
check.
Respondents from large scale industries (37.7%)
are mostly referred to the company clinic while the
less financially buoyant small scale industries refer
a sizeable (22.5%) number of their workers to the
government general hospitals. Fortunately, the
medical bills are settled partially or in full by the
companies depending on the amount.
The high percentage of referal available may be a
reflection of 75% awareness of health consequences
of industrial solid waste. On the frequency of
attendance of health and safety training programme,
respondents from large scale (71.1%) were better
exposed to contemporary health and safety
regulations and guideline even though the industries
have no health policy plan.
Conclusion
The results from the present study have indicated
that waste management from the industries studied
is inadequate. The situation is worst among small
scale industries. Companies need to be responsible
for their industrial waste management. The small
and large scale industries should be required to seek
authorization from Abia State Environmental
Protection agency under relevant rules; equally the
agency should be committed in enforcing the
.
relevant rules. Towards this, law should be
promulgated to tax the industries. The tax will help
offset the environmental damage by going towards
environmental restoration, protection and spreading
information to increase knowledge on these issues.
It is possible that through the efforts of ASEPA and
industries, a mechanism could be evolved for better
management.
The industries should be made to undertake the
detailed risk assessment of the waste. A policy
should be formulated based on reduce, reuse,
recover and dispose by the industries. Companies
are expected to improve their waste and
environmental performance and efficiency on a
regular basis.
Segregation should be done at the point of waste
generation and can be achieved through proper
training, cleaning standard and tough enforcement.
Industrial waste should be collected and transported
in safe containers. The open dump site should be
closed and municipal should work in conjunction
with the industrial sector towards development of
specific sanitary landfill for the treatment and
disposal of waste. Facility should be installed to
extract gas which can be burnt to generate the
needed electricity in Aba.
Personnel handling wastes of the industries
should be made to wear appropriate protective
clothing. Mechanical methods for handling waste
should be adopted where possible and people
should be educated about the dangers of manual
handling of hazardous waste. Recycling plants as in
developed countries should be built to convert some
of the waste to useable materials and become a
source of employment and income to the industries
and the society [29]. The uncontrolled incineration
should be discouraged, rather hygienic and well
structured incinerators should be established
outside the city to take care of the combustibles.
Generally the government should promote
environmental management system in all industries
through awareness programmes training and
demonstration projects which should cover all
stakeholders concerned. There should proper
budget allocation for waste management activities
and most importantly industries should be
encouraged to employ professionals.
________________________________________________________________________________
References
[1] PDE, (1994). Project in Development and the Environment. Comparing
Environmental Health Risks in Cairo, Vol. 2.
[2] Olanrewaju, A.O., (2000). Refuse: Tinubu’s Victory. P.M. News (Lagos), March
[3] Palczynski ,R. J(2002) Study on Solid Waste Management Options for Poverty
Africa.Project report African Development Bank Sustainable
Development & Reduction Unit
[4] Adekoya, N and Bishop C (1992). Failure of environmental health programme in Nigeria: A
.
concern for health planners in developing nation. International Journal of Environmental Health
Research 2: 167-170.
[5] Ajero,C.M.U. and Ukaga,C.N. (2006). Development, Global Change and Public Health.
Mwgasoft Owerri
[6]Ayotamuno, M.J., Godo, A.E. (2000). Municipal solid waste management in Port Harcourt,
Nigeria, obstacles and prospects. Management of environmental quality. African International
Journal 15(4):389-398
[7] Kalu, C, Modugu, W.W. and Ubochi, I. (2009). Evaluation of solid waste management policy in
Benin Metropolis, Edo State Nigeria. African Scientist 10 (1) 1595.
[8] Oyeniyi, B.A. (2011) Waste Management in Contemporary Nigeria. The Abuja Example.
International Journal of Politics and Good Governance 2 (2.2) Quarter 11 2011.
[9] Assessment on industrial waste management and it’s impact on environment
http://www.wisegcek.com what is-industrial- waste management. htm) retrived 12/2/2012
[10] Afon, A (2007). Analysis of solid waste generation in a traditional African city: the example of
Ogbomoso, Nigeria. http//eau.Sagepub.com/coutent/19/2/527.refs.html.Retrived 18/1/2012.
[11 ]Nabegu, A.B. (2010). Analysis of Municipal Solid Waste in Kano, Metropolis, Nigeria.
Journal of Human Ecology 31 (2): 111-119
[12] Federal Environmental Protection Agency(1991).Guidelines and Standard for Industrial
Effluents,Gaseous emission and Harzardous waste Management in nigeria.
[13] NPC (1992). National Population Commision, Surulere Lagos Nigeria.
[14] JICA (1998). The Study on Solid Waste Management in Nairobi City in The
Republic of Kenya. JICA Final Report, Nairobi City Council, August
[15] Diaz, L.(1999). An Overview of Solid Waste Management in Economically
Developing Countries.A Regional Conference on Integrated Solid Waste
Management. CapeTown, South Africa, October 25.
[16] Odochu,J.N.K.(1994) Waste Generation and Management in a Depressed Economy. 2nd
Edition, Geirge G.London
[17] United Nations Environmental Programme, UNEP (1991) Technical Report Series No. 7.
Audit and Reduction Manual for industrial Emissions and Wastes UNEP, Nairobi, Kenya.
Pp.26-32.
[18] United Nations environmental Programmers, UNEP (1995) Technical Support document on
residential and industrial (solid) waste disposal. UNEP, Nairobi ,Kenya
[19] Zurbrugg,C (1999).The chanllenge of solid waste disposal in developing countries, SANDEC
NEWS EAWAG No 4,1999.
[20] Cunningham, W.P., Cunningham,M.A. and Saigo, B.W.,(2003). Environmental Science: A
global concern. Mc-Graw Hill Book Companies, America.7th edition .
[21] Longe, C.O. Kehinde, M.O (2005). Investigation of Potential Groundwater Impacts on an
.
unlined Waste disposal site in Agege, Lagos, Nigeria. Proc. 3rd Faculty of Engineering
International Conference. University of Lagos Nigeria
[22] Hoornweg, B, Arnold, K and Smith, H (1999). Landfill gases as pollutant. A preliminary
report from Mamilla MAB Technical Paper No 14 Paris UNESCO.
[23] Agunwamba,J.C.(1998).Solid waste management in Nigeria-Problems and issues.
Environmental Management 22(6):849-856
[24] Agunwamba, J.C (2003). Analysis of scavengers’ activities and recycling in some cities of
Nigeria. Environmental Management 32 (1): 116-127
[25] NCC(2000). African Development Bank, Solid Waste Management
Questionnaire, Nairobi
City Council. Nairobi, Kenya.
[26] B a k u A (2010) . Working environment in some establishments in the UAE rural areas
.International Journal of Academic Research Vol. 2. No. 6. Part I
[27] Ec (1998). Proposal for a Council Directive on the Incineration of waste.
Brussels 07.10.1998 Com(1999) 558 final.98/0289 (SYN).
[28] Banjo A.D, Adebanibo, A.A.R and Dairo, O.S. (2011). Perception and operational habit of the
waste managers of domestic waste disposed in Ijebu ode, South East Nigeria. African journal of
Basic and Applied Science 3 (4): 131-135
[29] Dhussa, A.K and Varshney, A.K (2000). Energy Recovery from Municipal Solid WastePotential and possibility. Bioenergy News, UNDP, 4 (1): 18-21
.
Deploying Electronic Roadside Vehicle Identification Technology to Intercept Small
Arms and Ammunition on Nigeria Roads
Akaenyi, I.W+, Osuagwu O.E.*
+Computer Education Department, Federal College Of Education (Tech.)
Umunze, Anambra State. ikechukwuakaenyi@yahoo.com +2347032022241
*Department of Information Mgt. Technology, Federal University of Technology, Owerri.
drosuagwu@yahoo.com +23480371017932
Abstract
The challenge of insecurity of life and property in Nigeria has assumed a frightening dimension. The security
situation keeps degenerating daily in spite of government’s acclaimed effort to contain the situation. This
implies that Nigeria of today has a complex security management challenges to handle in order to liberate her
citizens from the bondage of insecurity of lives and property, ranging from kidnapping, armed robbery,
militancy, suicide bombing, ritual murders and human parts selling. The arms being used in perpetuating some
of these criminal acts are conveyed by the criminals using our roads. These security lapses are still possible in
spite of huge security personnel presence on Nigeria road checkpoints. This implies that the present system of
“stop and search” operation is defective and inefficient to handle current security dynamics. To ensure
effectiveness of roadside policing, there is need to carry out this “stop and search” operation using electronic
security system. The objective of this paper, therefore, is to present a model archetype that would be capable
of sweeping through commercial and private vehicles on the move automatically using Wireless Sensor
Networks, Vehicular Ad Hoc Networks (VANETS), OCR, transponders linking all sensitive security
observatories to a central data base for verifications, security alerts to the security agencies for prompt action
and national security control. The paper presents a data base network and communications architecture from a
roadside observatory through to the computer control room and then security personnel on duty. The present
practice of stop and search fails to capture most vehicles conveying sensitive and dangerous security exhibits
such as chemical, small and light arms. Moreover the present system is cumbersome, stressful, time consuming
thus reducing the desired reliability, accuracy of roadside policing.
Key Words: small arms and ammunitions, transponder, scanners. Sensors. RFID
______________________________________________________________________________
.
Introduction
Security is the degree of protection against
danger, damage, loss and criminal activities. It is
the state of being free from danger. It is the
measure taken as a precaution against theft,
espionage or sabotage [1]. In other words, security
must take into account the actions of people
attempting to cause destruction. Security as a form
of protection is a structure and processes that
provide or improve security as a condition. It
involves freedom from risk, or danger, doubt,
anxiety or fear. More so, security provides safety
and confidence in the people of any nation as well
as foreigners. However, perception of security by
criminals will deter malicious behaviour especially
with visual signs of security protections such as
video surveillance, alarm systems, closed circuit
television (CCTV) and other automated Roadside
Vehicle monitoring. This is physical security that
describes measures that prevent or deter criminal
attackers from accessing a facility illegally. These
monitoring technologies assist in fortifying national
security.
However, it is very important to be secured and
so all security policies must be enforced by
mechanisms that are strong enough with sense of
commitment. There are organized methodologies
and risk assessment strategies to assure
completeness of security policies and ensuring that
they are completely enforced. Hence, security
enforcement is very vital for the economic
development of any nation like Nigeria. It is when
this is assured that humans and goods can be easily
transported from one location to another. To have
effective security in the nation, there is need to
check physically the content of any vehicle on
transit. The present “stop and search” operation
done manually to track arms and ammunition on
Nigeria roads has not achieved the desired security
objective. It is then necessary to use electronic
means to do actualize this goal. This will help in
tracking down small arms and ammunition being
conveyed by vehicle since road is the major means
of transportation in Nigeria.
Arms
Small arms is a term used by arms forces to
denote infantry weapon an individual soldier may
carry [2]. The description is usually limited to
revolvers, pistols, sub-machine guns, carbines
assault rifles, rifles, sniper rifles, squad automatic
weapons, high machine guns, and sometimes hand
grenades. Short guns, general purpose machine
guns, medium machine guns, and grenade launchers
may be considered small arms or as support
weapons, depending on the branch of the armed
forces. Small arms typically do not include infantry
support weapons. In the U.S military, small arms
refer to hand guns or other firearms less than 20mm
in caliber, and including machine guns [3]. The
North Atlantic Treaty Organization (NATO)
definition in [4], extends to “all crew-portable
direct fire weapons of a caliber less than 50mm and
will include a secondary capability to defeat light
armour and helicopters”. Though there is no
civilian definition of small arms, but the term
encompasses both small arms and light weapons.
Ammunition
Ammunition is a generic term derived from the
French Language La munitions which embraced all
material used for war, but which in time came to
refer specifically to gun powder and artillery. The
collective term for all types of ammunition is
.
munitions [5]. In the widest sense of the word it
covers anything that can be used in combat that
includes bombs, missiles, warheads and mines. The
purpose of ammunition is predominantly to project
force against a selected target. However, since the
design of the cartridge, the meaning has been
transferred to the assembly of a projectile and its
propellant in a single package.
Current Security Scenario in Nigeria
For some time now, the issue of insecurity of life
and property in Nigeria has assumed a frightening
dimension.
The
security situation
keeps
degenerating day by day and getting sophisticated
with every Sunday bombing. This implies that
Nigeria of today has a
complex security
management challenge to handle in order to liberate
her citizens and foreigners alike from the bondage
of insecurity of lives and property, ranging from
kidnapping, religious disturbances, political
gangsters, armed robbery, militancy, suicide
bombing, arms proliferation and movement.
Nigeria is a developing democratic country in
West Africa and the most populous black nation in
the world. It has the largest population on the
continent of Africa. The population is estimated at
149 million people. However, the country has
experienced a number of crimes, political, religious
and cultural instabilities. This usually involves
violence, destruction of properties and human life
using arms and ammunition. One tends to wonder
how these weapons of mass destruction reach the
civilians in the village that enable them to use it at
the slightest provocation. This demonstates that the
security situation of this great country is in poor
management state and therefore requires retooling.
Radio Nigeria reported on Wednesday, 18th
August, 2010 [6] in its seven o’clock network news
in the morning, that fire arms were discovered in a
504 Peugeot saloon car, traveling from Maiduguri
to Kaduna and that these fire arms were neatly
parked beneath the car’s seats. This was discovered
at a police checkpoint. It is only a case out of many
cars having such arms passing through the
checkpoints in Nigeria undetected. It further
highlights that most of such arms and ammunition
are transported through road using vehicles which
is the major means of transportation in Nigeria.
On 15th June, 2010, the Department of State of
USA, [7] warned U.S. citizens through a document
titled “Travel Warning for Nigeria” circulated
through the net that it is highly risky to travel to
Nigeria. The Department of State recommended
that while in the country the US citizens should
avoid all but essential travel to some states of
Nigeria. This shows how the international
communities are seeing the security situation in the
country. Furthermore, the document came up
because of high rate of crime in the country carried
out with arms and ammunition. The crimes includes
kidnapping, armed robbery, suicide bombing and
other armed attacks. Crimes in Nigeria are done by
individual criminals and gangs, as well as some
persons wearing police and military uniforms.
In Nigeria, the citizens and foreigners have
experienced armed muggings, assaults, burglary,
carjacking, rape, kidnapping and extortions
involving violence. Home invasions remain a
serious threat, with armed robbers accessing even
guarded compounds by scaling perimeter walls,
following or tailgating residents or visitors arriving
by car into the compound, subdoing guards and
gaining entry into homes or apartments. However,
.
Nigerians and expatriates have been victims of
armed robbery at banks, gas stations, grocery stores
and on airport roads during daylight or evening
hours. These criminal acts are performed by people
carrying arms and ammunition in their vehicles
passing law enforcement checkpoints unnoticed and
undetected of the arms and ammunition. This is a
big security problem throughout the country. It
gives a sense of insecurity which casts an aspersion
on the Nigeria security system.
officers do not render twenty fours service. This is a
serious breach on security system of this country.
However, the advantages in this present system are
that:
• It is cheaper since no special instrument is
acquired in carrying out the operations apart from
the usual gun.
• It does not involve special training of security
personnel apart from the usual training given to
them at the Police College.
Present System of Tracking Arms and
Ammunition on Nigeria Roads
Presently, the detection and tracking of arms and
ammunition on Nigeria roads is done manually,
through what t Nigeria Police call “stop and
search” operation. In most cases the police and
other law enforcement officers will block the road
thereby forcing motorists to stop at the check
points. The officers will then approach the stopped
vehicle and ask the drivers what they are carrying
in their vehicle boots and then peep into the vehicle
to observe contents. The officer may ask the driver
to come down and open the boot to enable him see
the content of the boot. When opened, the officer
will usually look into the boot without searching
and then may ask the vehicle to go. In some
situations, proper searching of the vehicle will be
done while some will not be searched at all.
Government vehicles, politicians’ vehicles and rich
men’s vehicles are not usually stopped for
searching. Motorists who give, the law enforcement
officers money are sometimes not searched.
Motorists having any arms or ammunition usually
never stop at checkpoints. Sometimes they wait
until the law enforcement officers have left the
checkpoints to pass since the law enforcement
Weaknesses of Present System of Tracking Arms
and Ammunition on Nigeria Roads
From the analysis of present system of tracking
arms and ammunition on Nigeria roads, it is very
obvious that the security operations at Nigeria
checkpoint is very porous and ineffective in
tracking arms and ammunitions. The “stop and
search” operations carried out on vehicles at
checkpoint is done manually. This approach isly it
is very stressful, cumbersome, and involves
tiredness on the part of the security agents and road
users. It also creates traffic jams and makes
travelling very uncomfortable. At this stage, the
security agents usually allow some vehicle to pass
unchecked. This implies that some vehicles with
dangerous arms and ammunitions can be allowed to
pass safely undetected. On the part of the road
users, they become unnecessary aggressive to the
security agents and some will start disobeying the
security agents. Some military officers, police
officers, politicians etc pass the checkpoint without
being stopped to be searched. How are we sure that
they are not having illegal arms or ammunition with
them?
When the “stop and search” operation is being
done, most of the road users are impatient with
.
security agents. They will start to drive recklessly
which usually end up with avoidable motor
accidents leading to some deaths of people.
However, the stop and search” operation goes with
traffic jam which is usually not acceptable to the
road users. At times the operation of “stop and
search” is not properly done due to bribery and
corruption or not done at all at the checkpoints.
Some security agents at checkpoint collect money
from motorists and other road users without
searching their vehicles which has negative
implications to the security of the nation. Refusal to
offer bribery may lead to killing of innocent
citizens by the security operatives. This was the
case on Saturday, 13th August 2011 at Nwagu
Junction, Agulu, Anambra State, where a 26-year
old man was shut dead by a police man because he
refused to offer bribe of twenty naira [8].
These weaknesses are very dangerous to the
security of the entire nation. Although there are
cases of getting these arms and ammunition at
checkpoints, the number passed undetected is
believed to be higher since road is the major means
of transportation in Nigeria. These problems and
more can be avoided if automated road side vehicle
identification system is deployed at the road sides.
This will track down any arms or ammunition being
transported through our roads.
Proposed Technology Solution
In the new system, introducing the use of
electronic road side vehicle identification system,
for tracking small arms and ammunition movement
at the checkpoint, will contribute positively to the
security of the nation. The “stop and search”
operation that is done manually by security agents
will be carried out automatically by this device. The
device will be mounted at the checkpoint having a
transponder fixed overhead. It will use the overhead
transponder to scan the approaching vehicle which
will also have transponder fixed on it. On reaching
the checkpoint the vehicle will slow down, while
the transponder connected with the electronic road
side vehicle identification system (RSVIS) will
scan the vehicle. The image and vehicle content
will be transmitted to the computer monitoring
screen kept beside the road in a house. If the
vehicles has no metallic arms the vehicles continues
his movement with green light showing. But if the
vehicle is having any arms or ammunition the
image of the arm will be shown on the monitoring
screen and at the same time blow alarm with red
light blinking or showing. The security officers at
the checkpoint will then park the vehicle and
properly search the vehicle now. The vehicle
particulars, driver’s personal data, the vehicle
owner’s data, description of the weapon will be
obtained and typed into the database through the
computer system attached with the RSVIS. After
taking these data the security officer hands over the
vehicle, driver etc to the police for further
investigation and prosecution.
In-dept analysis of Nigeria road transportation
system have shown that it is possible to deploy
electronic RSVIS for tracking small arms and
ammunition movement on Nigeria roads. This
system will remove or minimize human
inadequacies affecting negatively the manual
searching of vehicles at checkpoints. It will ensure
and introduce fairness, fastness, objectivity and
accuracy up to eighty percent in checkpoint
searching thereby reducing violent crime through
proactive measures.
Overall Data Flow Diagram of the New System: The overall data flow diagram (ODFD) of the
Proposed new system is shown below:
Vehicle moving
towards the
scanning point
OCR/Scanner
screening
Vehicle content
identified
If no
If weapon
weapon
Vehicle
continues
RVIS show
the content on
Security
officer stops
Raise alert to
draw attention
Security officer
takes data using
Security
officer hands
over the
vehicle
Nigeria Police
Further Investigation
Discharge and
Court Judgment
Prosecute in
Imprisonment
Figure 1.1: ODFD of the new system
9
.
The Flowchart of the New System: This is shown below
We propose here for adaptation a hybrid ITS
safety architecture that combines vehicle-to-vehicle
communication and vehicle-to-roadside sensor
communication [9]. Opposed to dedicated roadside
units, which require major investments for
purchase, installation and maintenance, roadside
wireless sensor and networking technology
represents a cost-effective solution and can leverage
the deployment of the system as a whole. Among
the various services of the hybrid communication
system proposed here include accident prevention
and post-accident investigation. Presented here also
is a system and protocol architecture with a fully
.
distributed concept for efficient and secure storage
of sensor data. For deployment, this architecture
will likely be combined with an alternative
approach using dedicated road-side units as a
centralized network element for communication
and data storage. For the proposed system, we
describe the main components (radio, networking
and services, security). Finally, we describe our
prototype implementation and experimental tested
featuring hardware and software platforms for
vehicle on-board units and sensor nodes. Figure 1.3
presents the proposed WSN architecture with
distributed
and
centralized
data
storage
Fig. 1.3: Systems architecture with distributed and centralized Data Storage[9][11]
The proposed Main technology components of the architecture are:
(i) radio interfacesIEEE 802.11p and IEEE 802.15.4,
(ii) routing protocols Geocast and tinyLUNAR,
(iii) middleware for VANETs and tiny PEDS for WSNs, and
(iv) Applications.
The components are well adapted to the specific requirements of VANETs and WSN, respectively.
VANET = Vehicular Ad hoc Networks
WSN = Wireless Sensor Networks
1
.
Table 1.1: Prototype Platforms For VANET And WSN Nodes [11]
Technology Integration Components.
These are technological equipment needed or to
be used in carrying out the tracking of metal arms
in vehicles which will work with RVIS.
( i)Transponder
This is a wireless communication, monitoring
and control device that picks up and automatically
responds to an incoming signal. It is a contraction
of the words transmitter and responder. Simply
put, transponder is an electric device used to
wirelessly receive and transmit electrical signals
[9]. It is an electronic device that can be used to
send and receive signals wirelessly.
Transponders were developed to be attached to
objects which needed to be located. A transponder
functions by receiving a signal, called an
“interrogator” because it is effectively “asking” for
information, then automatically conveying a radio
wave at predetermined frequency. In order to
broadcast a signal on a different frequency than the
one received, a frequency converter is built in. By
receiving and transmitting on different frequencies,
the interrogator and transponder signals can be
detected simultaneously. In a real situation,
transponder will be attached to vehicles for easy
location and searching. This can be mounted on the
windshield or dashboard.
Transponders can be classified as active and
passive transponders. Active transponder is a type
of transponder employed in location, identification,
and navigation systems for commercial and private
vehicles. It transmits a coded signal when it
receives a request from a monitoring or control
point. The transponder output signal is tracked, so
the position of the transponder (on the vehicle) can
be constantly monitored. It will operate over a
kilometre distance and an example of this is Radio
Frequency
Identification
(RFID).
Passive
transponder is a type of transponder that allows a
computer or robot to identify an object. It must be
used with an active sensor that decodes and
.
transcribes the data the transponder contains.
store items are common examples. This is the one
to be employed in this work
Figure 1.4: Transponder over a truck
[Source: 9]
(ii) RFID
This acronym stands for Radio Frequency
Identification. It is the use of a wireless non-contact
system that uses radio frequency electromagnetic
fields to transfer data from a tag attached to an
object, for the purposes of automatic identification
and tracking. It is a technology that incorporates the
use of electromagnetic or electrostatic compiling in
Magnetic labels, such as those on credit cards, and.
the radio frequency portion of the electromagnetic
spectrum to uniquely identify an object, animal or
person. It does not require direct contact or line-ofsignal scanning.
A RFID system uses tags, or labels attached to
the objects to be identified. Two way radio
transmitter-receivers called interrogators or readers
send a signal to the tag and read its response. The
readers generally transmit their observations to a
computer system running RFID software or RFID
middle ware. The tag’s information is stored
electronically in a non-volatile memory. An RFID
reader transmits an encoded radio signal to
interrogate the tag. The tag receives the message
and responds with its identification information.
This may be only a unique tag serial number, or
may be product related information such as stock
number, lot or batch number, production date, or
other specific information. Note that RFID tags can
be either passive, active, or battery assisted passive.
However, RFID system consists of these three
components, antenna, transceiver (often combined
into one reader) and a transponder [9].
Figure 1.5: RFID [source: 2]
.
(iii)
Optical Character Recognition
This is often abbreviated as OCR, it refers to the
branch of computer science that involves reading
text from paper and translating the images into a
form that the computer can manipulate. It is the
recognition of printed or written text and characters
by a computer system or OCR reader/OCR
machine. It involves photo scanning of the text
character-by-character, analysis of the scanned-inimage, to identify each alphabetic letter or numeric
digit, and then the translation of the character image
into character codes such as ASCII, commonly used
in data processing. Special software is used to
compare the pattern of the signal sent from the
machine or OCR reader to patterns already stored in
the memory. When the match has been made, a
respond signal will be sent to the machine accepting
or rejecting the pattern. OCR is used in reading
vehicles number plates. However, OCR systems
include an optical scanner for reading text, and
sophisticated software for analyzing images, and
software
to
recognize
characters.
Figure 1.6: OCR[9]
(iv) Sensor
A sensor is a converter that measures a physical
quantity and converts it into a signal which can be
read by an observer or by an instrument which are
mostly now electronic. Generally, sensor is a device
that receives signal and converts it into electrical
form which can be further used for electronic
devices. For accuracy, most sensors are calibrated
against known standards.
In this work we are concerned with image sensor
which is a device that converts an optical image
into an electronic signal. It is used mostly in digital
cameras and other imaging devices. Most currently
used sensors are digital charge-coupled device
.
(CCD)
or
complementary
metal-oxidesemiconductor which is active pixel sensors
(CMOS APS). Both types of sensors accomplish
the same task of capturing light and converting it
into electrical signals.
An active-pixel sensor (APS) is an image sensor
consisting of an integrated circuit containing an
array of pixel sensors, each pixel containing a
photodetector and an active amplifier. There are
many types of active pixel sensors including the
CMOS APS used most commonly in cell phones
cameras, web cameras. Such an image sensor is
produced by a CMOS process (and is hence also
known as a CMOS sensor) and has emerged as an
alternative to charge couple device (CCD) image
sensors.
Figure 1.7: Image sensor [2]
(v)
GPS: This is Global Positioning System
which is a spaced based satellite navigation system
that provides location and time information in all
weather, anywhere or near the Earth, where there is
an unobstructed line of sight to four or more GPS
satellites. It is freely accessible to anyone with a
GPS receiver. Modern vehicle tracking system
commonly uses technology for locating vehicle. It
can locate a vehicle anywhere on earth if it has
vehicle tracking system. It also allows drivers of car
and trucks to ascertain their location anywhere on
earth. Mores, vehicle information can be viewed on
electronic maps via the internet or specialized
software.
(vi) Simulation: This is the imitation of the
operation of a real word processor system over
time. The act of simulating something first requires
that a model be developed; this model represents
the key characteristics or behaviours of the selected
physical or abstract system or process. The model
represents the system itself, whereas the simulation
represents the operation of the system over time.
Simulation is used in many contexts, such as
simulation of technology for performance
optimization and for scientific modeling of natural
systems or human systems to gain insight into their
functioning. It can also be used to show the
eventual real effects of alternative conditions and
courses of action. Simulation is also used when the
real system cannot be engaged or it is being
designed but not yet built like in this thesis. In this
thesis we are concerned with computer simulation.
Computer simulation is a computer program that
attempts to simulate an abstract model of a
particular system. Simulation of a system is
represented as the running of the system’s model. It
can be used to explore and gain new insights into
new technology and to estimate the performance of
systems too complex for analytical solutions.
Computer simulations vary from computer
programs that run for a few minutes, to network
.
based groups of computers running for hours, to
ongoing simulations that run for days.
(vii) Computer System: A complete, working
computer is known as computer system. It includes
the computer along with any software and
peripheral devices that are necessary to make the
computer function. Every computer system requires
an operating system. But a computer is
programmable machine. The two principal
characteristics of a computer are;
• It responds to a specific set of instructions in a
well defined manner.
• It can execute a pre-recorded list of instructions
that is program
(viii) Image Scanner: In computing, an image
scanner often abbreviated to just scanner, is a
device that optically scans images, print text,
handwriting, or an object, and coverts it to a digital
image. Hand held scanners have evolved from text
to three dimensional (3D) scanners. A 3D scanner
is a device that analyses a real world object or
environment to collect data on its shape and
possibly its appearance. It is being used in different
areas of life including inspection. The purpose of
3D scanner is usually to create a point cloud of
geometric samples on the surface of the subject. It
is analogous to cameras. Like cameras, they have a
cone-like field of view and can only collect
information about surfaces that are not obscured. A
3D scanner collects distance information about
surfaces within its field of view. The “picture”
produced by 3D scanner describes the distance to a
surface at each point in the picture. This allows the
three dimensional position of each point in the
picture to be identified.
However, for most situations, a single scan will
not produce a complete mode to the subject.
Multiple scans, even hundreds, from many different
directions are usually required to obtain information
about all sides of the object. These scans have to be
brought in a common reference system, a process
that is usually called alignment or registration and
then merged to create a complete model. This
whole process, going from the single range map to
the whole model, is usually known as the 3D
scanning pipeline.
There are varieties of technologies for digitally
acquiring the shape of 3D object. A well
establishment classification divides them into two
types: contact and non-contact 3D scanners. Non
contact 3D scanners can be further divided into two
main categories, active and passive scanners. In this
thesis, we are concerned with non contact 3D
scanners.
Active scanners emit some kind of radiation or
light and detect its reflection in order to probe an
object or environment. The possible types of
emissions used include light, ultrasound or x-ray.
But passive scanners do not emit any kind of
radiation themselves, instead rely on detecting
reflected ambient radiation. Most scanners of this
type detect visible light because it is a readily
available ambient radiation. Other types of
radiation, such as infrared could also be used.
Passive methods can be very cheap, because in
most cases they do not need particular hardware but
simple digital cameras.
However, the anticipated roadside Vehicle
Identification Technology for implementation is
shown below but not drawn to to scale.
.
Figure 1.8: Proposed Roadside Vehicle identification layout [9].
How e-screen works
• As a participating vehicle approaches an
enforcement station, the vehicle’s transponder is
read by a roadside reader.
• A computer in the enforcement station looks
up the transponder in a database to make sure the
vehicle is in good standing with regard to
registration, safety record, and is not carrying any
weapon.
• If everything checks out, the truck is given a
green light on the transponder, indicating that the
truck may proceed on down the road.
Enforcement personnel are then free to focus
their attention on high-risk carriers.
Figure 1.9: Proposed Electronic Screening Software[9]
.
Justification of the new system
The new system ensures equity, fairness,
fastness, reliability, objectivity and accuracy up to
eighty percent than the “stop and search” operation
at Nigeria road checkpoints. It will completely take
care of the weaknesses observed in the present
manual system. The system will automatically
generate report or information on vehicles with
arms or ammunition and then draw the attention of
security operatives on duty. Moreover, every
vehicle must be screened automatically. There
would no further incidence of allowing vehicles to
pass without being searched due to human
inadequacies such as tiredness. If the new
technology is properly implemented, it will reduce
the rate of violent crimes, riot, agitation, etc in
Nigeria involving the use of arms and ammunition.
This will improve the security situation of this great
country.
Anticipated Benefits of the Proposed System
•
For government:
o
More effective and
efficient screening.
o
Make better use of
security personnel.
o
Focus enforcement
resources on the high-risk
carriers.
o
More consistent and
ubiquitous
enforcement
coverage
o
Monitor more routes at
more times
o
More difficult for non-
o
o
•
Compliant carriers to avoid
enforcement.
Improved safety
Improved homeland
security
For industry:
o Creates a more level playing field
o More difficult for anyone to shortcut the requirements.
o Everyone must play by the same
rules.
o Better revenue base – everyone
paying their fair share
o Streamlined
operations
at
enforcement locations
o Safe and legal trucks avoid stops and
delays.
o Fuel and time savings.
o Improved safety
o Unsafe operators taken off the road.
o Less need to stop trucks on shoulder
to perform checks.
Conclusion
This paper has identified the emerging complex
security scenario in Nigeria as being capable of
disintegrating the Nigerian state if not urgently
addressed.
We have proposed a technology
solution via the deployment of Electronic Roadside
Vehicle Monitoring Technology using Wireless
Sensor Networks, Transponders, OCR and dynamic
camera and automated vehicle checkpoints to track
incidence of small arms movement in Nigeria. The
objective is to arrest the unbridled proliferation of
small arms used in perpetuating murder, robbery
.
and bombings around Nigeria. The authors strongly
believe that urgent implementation of this new tool
by the Federal Government will go a long way in
staving Nigeria away from the threat of sovereign
collapse.
Recommendations
In line with the foregoing observations about the
emerging complexity of the security scenario in
Nigeria, we recommend thus:.
• That comprehensive automation of
databases for vehicle registration throughout the
country is initiated as a matter of urgency.
• That issuance of vehicle number plate
should be comprehensively done with automated
database system.
.
• Driving license issuance should be
automated with a distributed or centralized database
and security and access controlled from a central
point.
• Vehicle registration database, vehicle
number plate database and driving license issuance
database must be linked and connected to the
RSVIS when implemented for analysis.
• Database for offenders should be
developed so, that reports can be generated at
regular intervals when implemented.
• That there be established Emergency
Security Response where all reports of identified
arm movement on the road is reported for
immediate security action.
________________________________________________________________________________
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
Merriam-Webster, (2010). Online Dictionary. Retrieved from http: //www.merriamwebster.com/dictionary/ security on 22nd August, 2010.
Wikipedia, (2010) Small Arms Free Encyclopedia retrieved online from
http://en.wikipedia.org/wiki/small_arms on 25th October, 2010.
Merchant-Smith, C.J & Haslam,, P.R (1982). Small Arms and Cannons. London:
Brassey’s Publishers.
Dikshif, P (1994) Proliferation of Small Arms and Minor Weapons, strategic
Analysis Journal of Proliferation of Arms and Weapons Volume 1 (2)
Chisholm, H. (ed) (2000). Ammunition. Encyclopedia Britisannia, USA: Cambridge
University Press.
Radio Nigeria (2010). Seven O’clock Network Morning News of 18th August, 2010.
USA (2010) Travel Warning for Nigeria. Retrieved online from
.
http://travel.state.gov/travel on 22nd August, 2010. Published by Department of State of USS
on 15th June 2010.
[8]
Radio Nigeria (2011) Seven O’clock Network Morning News of 1st September, 2012.
[9]
Osuagwu, O.E.(2010) Deploying Electronic Roadside Vehicle Identification to Track
Transit Ammunition Movements: A Tool for Enhancing the Security of Nigerian State. A paper
Delivered at the International Conference of the Nigeria Computer Society, held at Asaba from
26th to 30th July, 2010.
[10] Dellaporta, J. (2010) What is Transponder? Retrieved online from
“http://www.wisegeek.com/” On 29th September, 2010.
[11] http://www.vanet.info/
.
Statistical Analysis of Deviance among Children in Makurdi Metropolis
Kembe, M.M* and Kembe, E.M+
*Department of Mathematics, Benue State University, Makurdi
+ Department of Home Science and Management, University of Agriculture, Makurdi
Abstract
This study sampled a total of four hundred and three individuals from designated households in Makurdi
metropolis, Benue State. The study respondents responded to a self-report survey which gathered information
on three deviant acts: alcoholism, smoking and dropping out of school. Criteria for deviant acts were defined,
and each of the three acts was analyzed against the criteria of the type of family. Propensities toward deviance
were identified against three normative dimensions – single parent (mother), single parent (father), both
parents. The most prevalent deviance among children from single parents is children dropping out of school
while that of children from two parents is alcoholism. It is the conclusion of this study that home background
and family type has a correlation with deviant behavior. Based on this, there is need for youth employment,
government intervention and awareness education on parental responsibility.
Keywords: Households, deviance, drooping out of school, smoking and alcoholism
1.0
Introduction
Deviance is behaving contrary to acceptable
norms and expectation of a society. Deviant
behavior is common with young children and
adolescent especially in their formative years when
character is formed and reformed to suit the
expectation of the family and the community. Every
society has specific behavioral standards and ways
in which people are supposed to act; sometimes
these are the paradigms for predictable behavior in
the society.
Deviance is a departure of certain types of
behavior from the norms of a particular society at a
particular time and place. In Nigeria, deviance is
described as a violation of culturally acceptable
norms and or a failure to conform to set rules and
ways of doing something that is traditionally
prescribed.
According to Santrock [10], many parents and
teachers often wonder about the rebelliousness
(truancy, anti–social behavior, disrespect for
constituted authority, sexual harassment, rape,
arson, destruction, adolescents delinquency) that
are portrayed by children and adolescent.
A behavior considered as deviant in one society
may be seen as non-deviant in another society. For
example, the traditional African social custom
appreciates chastity, modest dressing, good morals,
.
decent behavior, and respect for elders, hard work
and integrity and frowns at premarital and extra
marital sexual relationship. It also prohibits
marriage between same sex such as homosexuality
and consanguine sexual relationship. The Nigerian
society frowns at alternative marriage styles and
parenthood, for example, single parenthood and
cohabitation. In some other societies, these
unacceptable behaviors are acceptable. That is why;
deviance is relative to time and place.
Studies conducted by Mallum [9] and Kembe [7]
shows that the home background plays an important
role in the child’s personality development. The
home is often critical to the type of behavior
displayed in the society. The traditional family as
reported by Shindi [11] is made up of family
members who would normally contribute to the
upbringing of the child. The extended family
system provided a wide range of stimulating
interactions that positively nurture the child,
particularly in the psychosocial terms.
In the absence of parental support provided by
both parents and the social support network
provided by the extended family, there are bound to
be problems in child rearing and upbringing.
According to Karst [6], a Single parent is a parent
who cares for children without the assistance of the
other parent in the home. Child birth and child
rearing is a task that involves the responsibility of
both parents.
The number of one parent families has increased
substantially. According to the United States
Census Bureau [9], about 20 million children live in
household which are single-headed. In the 1970s,
divorce was a common reason for single parenting,
and during this period the number of families
headed by one parent increased rapidly. The
number peaked in the 1980s and the declined
slightly in the 1990s. By 1996, 31 percent of
children lived in single parent families.
Apart from divorce, single teenage parenting is
also reported among adolescent. This occurs as a
result of teenage unwanted pregnancy leading to
school dropout and other health consequences [4].
Furthermore, it has been proved that single headed
families are not economically stable. Children from
economic dependent homes are likely prey to social
vices like stealing and school truancy. Indeed, the
incidences of child abuse (street hawking, house
maid, bus conductors, mine, factory worker, child
prostitution) are as a result of poverty. Poverty has
also forced some parents to abandon their primary
responsibility of child care and nurturance. This in
turn has resulted in all kinds of delinquent behavior
such as compulsive television viewing, drug
addiction, cultism and alcoholism.
Farrington [3] reports that 90% of adolescent
boys and girls in intact families were within the
normal range on behavioral problems.10% had
serious problems that we would generally require
some type of professional help. The percentages for
divorced families’ were 74% of the boys and 66%
of the girls in the normal range and 26% of the boys
and 34% of the girls were in the problematic range.
It should be noted that reasons for single
parenthood can be as a result of death of spouse.
Spousal death of either the mother or the father can
result in child rearing by only one parent. Normally,
the family suffers set back especially if it is the
bread winner of the family. Deviance behavior can
result from the psychosocial loss of the loved one.
So, there may be levels of deviance depending on
the causal- effect relationship.
.
Recently, research in social science has attracted
attention on the effects of mass media on violent
and aggressive behavior. According to Anderson,
Berkowitz, Donnerstein, Huesman, Johnson, Linz,
Malamuth and Wartella [1] provided converging
evidence linking frequent exposure to violent media
in childhood with aggression later in life, including
physical assault, alcoholism and other forms of
abusive behavior.
2.0
Statement of the Problem
Deviance has been reported variously as behavior
that is contrary to acceptable cultural norms of a
society. In Benue state recently, there is a ban on
the consumption of illicit gin especially by the
youths. The reason for the ban on drinking is not far
from the deviant behavior that is exhibited by the
youths as seen in hooliganism, illicit sexual
activities, rape and other forms of abusive and antisocial behavior.
It is also a deviant behavior to see teenage girls
who are supposed to be in school to be engaged in
pregnancy and child rearing. Teenage pregnancy is
often the reason for school dropout among
adolescent girls [10], [4] . 40% of teenage mothers
drop out of school to rear their children as single
parents. Single parenting places the burden and
responsibility of intact families only on one parent.
This raises doubts on the psycho social behavior of
these children. Society is not oblivious of the fact
that home background affects the behavior of
children,
[9],[1]. It is in view of this discourse, that this study
intends to look at the deviant behavior exhibited by
children from single parenthood and see whether
there is a relationship between these behavior and
those of children from intact families.
Methods
The research work was carried out in Makurdi,
the capital of Benue State.
A structured
questionnaire was developed and given to our
targeted respondents. A survey was sent to a
randomly selected 500 Households. Of the 500
Households, 403 returned the questionnaire (81%).
Analyses for this study include descriptive statistics
and non parametric methods and the level of
significance was set at 0.05.
Results and Discussions
What are the different family status in Makurdi Metropolis?
Table 1. Percentage response on types of Family status in Makurdi
Family status
Responses
(f)
(%)
Single parent(mother)
220
54.6
Single parent (father)
102
25.3
Two parents(mother and father)
81
20.1
Total
403
100
Result from Table 1 shows that majority of families
in Makurdi metropolis are composed of single
parents who are women (55%), other types of
.
family status include single parents made up of
men(25%) and two parents(20%).
What are the prevalent deviances in Makurdi
Metropolis
Table 2. Prevalence of deviance in Makurdi Metropolis
Deviances
Alcoholics
Smoking
Dropping out of School/runaway
Teen parents
Cultism
Political thuggery
Robbery(petty robbery, armed robbery and
house breaking)
Commit Suicide
Total
The result in Table 2 shows that 24.8% indicated
that the most prevalent deviance amongst children
in Makurdi town is alcohol. This is closely
followed by smoking which is represented by
16.9%. Dropping out of school and run away from
home is another deviance that is on the rise
representing 16.1%. Cultism is also an emerging
problem on the list of deviance in Makurdi town
representing 13.6%.
Responses
(f)
100
68
65
43
55
52
18
2
403
(%)
24.8
16.9
16.1
10.7
13.6
12.9
4.5
0.5
100
Cumulative Deviance of Children from Single
Parents
Figure 1 depicts the cumulative sums of
deviances in children from single parents. A plot of
these cumulative deviances is presented in figure 1
below. It shows that the number of drop outs from
school was highest followed by those taking
alcohol.
.
Figure1. Cumulative sums of the three components of deviance in children of single parents.
Cumulative Deviances of Children from Two
Parent Families
Figure 2 depicts the cumulative sums of
deviances in children from two parents. A plot of
these cumulative deviances presented in the figure
below shows that the number engaged in alcohol
was highest and this is closely followed by the
number of smokers and those dropping out of
school the least.
Figure 2. Cumulative sums of the three components of deviance in children of two parents.
.
Comparison of deviance of children from single
parents with those from two parents
Figure 3 depicts cumulative deviance from
children brought up from single parents compared
with those brought from two parents. It clearly
shows that the highest deviance is exhibited by
children from single parents than children from two
parents.
Figure 3. Cumulative deviances from two types of family background
Hypothesis
H0: There is no significant difference between
family status and deviant behavior of children. The
Chi- Square ( ) statistics was used in the test of
independence.
Table 3. The relationship between the family type and the observed defiant behavior of children.
Parent status Alcoholics
Smokers
School
dropouts
Single
20
19
18
Parent(Mother)
Single
48
70
68
Parent(Father)
Parents(Mother 54
76
20
and Father)
Total
122
165
116
The null hypothesis:
j=1,2,…,J
i=1,2…I;
Alternate Hypothesis:
Test Statistic;
.
the proportion of individuals in the
Where
population who belong in category I of factor 1 and
category j of factor 2
= P (a randomly selected individual
falls in both category i of factor 1 and category j of
factor 2
= P( a randomly selected individual
falls in category i of factor 1)
= P( a randomly selected individual
falls in category j of factor 2)
Rejection region:
Thus
But
Since
,
We therefore reject hypothesis of independence.
We conclude that the status of parents of children
does give information about the deviant behaviour
of children. In particular Children who are
alcoholics appear more likely to come from a single
parent that is a mother.
Discussions
Results from analysis show that there are more
single parents who are mothers in Makurdi
metropolis. The vulnerability of women to poverty
and the imbalances in education and employment
coupled with the responsibility of housekeeping and
child rearing affects the family. Recently singleheaded families have become rampant either from
the male-headed or female-headed households,
leaving the burden of housekeeping and childrearing a challenge. There have been various
reasons given for the types of family status, ranging
from divorce, death of spouse, lifestyle and other
social circumstances including out of wedlock
births.
According to Kandel [5], Social scientists have
found that children growing up in single-parent
families are disadvantaged in other ways when
compared to a two parent families. Many of these
problems are directly related to the poor economic
condition of single-parent families, not just to
parenting style. These children are at risk for the
following: lower levels of educational achievement
likely from irregular school attendance
i.
likely to drop out of school
resulting from unmet needs
ii.
likely to become teen parents
and face other social hazards of
teenage pregnancy
iii.
likely to become thugs and be
involved in “boko haram”
iv.
frequently abuse drugs and
alcohol
v.
high-risk sexual behavior
vi.
likely to engage in cultism
vii.
likely to commit suicide
According to Kembe (2005), a child is an
ambassador of a home, thus whatever behavior the
child exhibits both in the school and in the larger
society is a direct function of how the child was
brought up with some few exceptions though.
Conger and Mussen [2] stated that children
behave differently according to their home type.
Home background plays a significant and critical
role in the overall personality output of individuals.
Children life adjustment is influenced from the
home background, for example, children who have
.
lived in deprived and or abusive environment are
most likely to exhibit negative societal behaviours.
Also, children from two-parent structured homes
are less likely to be engaged in deviances like
dropping out of school, smoking, alcoholism
compared to the children who are from
unconventional homes.
Conclusion and Recommendations
The overall result of this study suggests that
children from single parents homes, on average, are
more engaged in deviances than children in two
parent families. The common forms of deviant
behaviors are school dropout, alcoholism and
smoking. This has implication for later life
adjustments. It is the conclusion of this study that
home background and family type has a correlation
with deviant behavior. Based on this, there is need
for youth employment, government intervention
and parental responsibility.
In the light of t foregoing findings of this
research, the following recommendations are made.
Youth development
Youth unemployment resulting from lack of
appropriate skills for work will continue to
constitute a problem unless skill centres and
programmes are developed to engage youths and
impart skills especially for immediate small scale
business. This recommendation, if implemented,
will also improve the economic base of most
homes.
Government intervention
Government can place policies and laws that will
prosecute certain repeated cases of youth deviant
behavior such as housebreaking, cultism, smoking
of illicit drugs and alcoholism. The present law in
the Benue State prohibiting the sale and drinking of
alcohol in social events should be enforced and not
made a mockery of.
Parental Responsibility
Parents are responsible for the upkeep of their
families and to ensure that children grow up into
acceptable personalities for the society. Quality
family time, discipline and positive parenting style
have implication for controlling deviant behavior.
References
[1]
[2]
[3]
[4]
Anderson, C.A., et al (2003). The infulence of Media Violence on Youth Psychological
Science in the Public Interest, 4, 81-110.
Conger, J.J and Mussen, P.H (1997). Child Development and Personality. Harper & Row,
publishers, New York
Farrington, D.P. (1995). The Development of Offending and Antisocial Behavior from
Childhood: Key Findings from the Cambridge Study in Delinquent Development. Journal of
Child Psycho Psychiatry 36, 29-64
Gipson, J. D., Koenig, M. A. and Hindin, M. J. (2008),
The Effects of Unintended
Pregnancy on Infant, Child, and Parental Health: Studies in Family Planning. Vol 39(1), 18-38
.
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
Kandel , D. B. (1990).
Parenting Styles, Drug Use, and Children’s Adjustment.
.Journal of Marriage and family. Vol.52,(1),183-196
Karst,P.(2000).
The Single Mother's Survival Guide. Freedom, CA: Crossing Press
Kembe, M.M.(2005). Influence of Family Background on Pattern of Child Misbehavior in
Makurdi Metropolis. Journal of Home Economics Research, Vol. 6(10), 166-170.
Kembe, M.M.(2008). Verbal Abuse: Causes and effects on Adolescent personality. The
Journal of Family Development Vol 3(1), 14-21
Mallum, J.O. (2002). The Home Environment and the Educational development of the
child. Journal of the Nigerian society for educational Psychologists Vol 1(1), 40-48
Santrock, J.W. (2005). Adolescence. Boston: Mc Graw Hill
Shindi, J.(1989).
The Nigerian Child. An Introduction to Child Care. Plateau state:
welfare committee
United States Census Bureau (2002).
APPENDIX A
Cumulative sums of the three components of deviance in children of single parents.
S/N
alcoholics
Cm
smokers
Cm
school dropouts
Cm
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
1
2
1
1
2
1
3
0
0
2
3
0
1
3
1
2
2
5
3
2
1
3
4
5
7
8
11
11
11
13
16
16
17
20
21
23
24
29
32
34
0
1
0
4
0
3
2
0
0
3
0
2
1
2
0
1
0
1
2
1
0
1
1
5
5
8
10
10
10
13
13
15
16
18
18
19
19
20
22
23
4
5
2
1
0
2
4
1
1
2
2
3
5
3
3
2
2
1
1
2
4
9
11
12
12
14
18
19
19
21
23
26
31
34
37
39
41
42
43
45
.
Appendix B
Cumulative sums of the three components of deviance in children of two parent families
.
Appendix B
Cumulative sums of the three components of deviance in children of two parent families
S/N
Alcoholics
C
Smokers
C
School dropouts
C
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
2
3
2
0
4
2
2
4
3
2
3
1
0
2
4
0
1
3
3
2
2
5
7
7
11
13
15
19
22
24
27
28
28
30
34
34
35
38
41
43
0
1
0
0
1
2
1
3
1
2
1
2
3
0
1
0
3
1
3
1
0
1
1
1
2
4
5
8
9
11
12
14
17
17
18
18
21
22
25
26
0
0
2
0
1
1
1
1
0
0
2
0
1
1
0
0
O
1
1
0
0
0
2
2
3
4
4
6
6
6
8
8
9
10
10
10
10
11
12
12
.
Appendix C
Cumulative deviances from children brought up from single parents compared with those brought from two
parents
S/No
1
Cumulative deviance from single parents
5
Cumulative deviances from two parents
2
2
13
6
3
16
10
4
22
10
5
24
16
6
30
21
7
39
25
8
40
33
9
41
37
10
48
41
11
53
47
12
58
50
13
65
54
14
73
57
15
77
62
16
82
62
17
86
66
18
93
72
19
99
79
20
104
82
.
A Profile Analysis on the Effectiveness of Two kinds of Feeds on Poultry Birds.
Onuoha, Desmond O+. and Opara Pius N*.
+Department of Mathematics and Statistics, Federal Polytechnic Nekede, Owerri, Imo State.
E-mail: desonuoha@yahoo.com
*Department of Statistics,Imo State University Owerri
Abstract:
This study was carried out to find the effect of two types of poultry feeds on the weight of poultry birds using
chicken from Michika Farm in Igwuruta in Port Harcourt, Rivers State as a case study. The data was collected
for a period of eight weeks, where a sample size of sixty chickens was divided into two equal populations. The
data for the analysis was collected as a primary data through the method of direct observation and
measurement. The data were analyzed using hostelling T2 distribution, F-distribution to Test for parallel
profile. At the end of the analysis, it was found that the profile was not parallel. This shows that the levels of the
treatment on feeds are not the same.
Key words: poultry feeds, poultry birds, primary data, direct observation, hostelling T2 distribution,
F-distribution
_______________________________________________________________________________
1.0
Introduction
Poultry farming is one of most lucrative business
ventures one can embark upon if properly managed.
The management of poultry could be attributed to
the production of healthy and weighty birds in order
to maximize profit. To actualize this, one has to
adopt the best poultry feed on the birds. This work
is aimed at using Profile analysis to select the best
feeds needed for the poultry birds. In this case,
profile analysis could be described as a situation
where a number of treatments are administered to
two or more populations. [22] ,[15]
stated that the responses must be expressed in
similar unit and are assumed to be independent of
one another, for different populations.
Furthermore, in comparing two or more
populations, we might be faced with questions such
as; Is the population’s mean-vector the same? That
is µ1 = µ2.
By applying profile analysis to test for the
effectiveness of two different types of feeds on
poultry birds, the questions are formulated in a
stage-wise approach:
1.
Are the profiles parallel?
.
Or Equivalent: H01: µ1i - µ1(i-1) = µ2i - µ2(i-1) = µ3i µ3(i-1) . . . = µgi - µg(i-1);
i = 1, 2, …,p
2.
Assuming the profiles are parallel, are they
coincident?
Or Equivalent: H02: µ1i = µ2i = µ3i = . . . = µgi ; i =
1, 2, …,p
Sources and Method of Data Collection
The data for the analysis is a primary data
collected through the method of direct observation.
The method of direct observation entails observing
and recording events as it is happening. The data
was collected from an experiment conducted for a
period of eight (8) weeks and two days, with sixty
(60) chickens divided into two (2) equal parts. The
first two days of the measurements were not used to
get the real effect of feed. It has been observed by
Johnson et al (1986) that when a new diet
formulation is introduced or a new type of feed is
presented, birds will often refuse to eat for a period
of time or intake is reduced. Group A made up of
thirty (30) chickens, were subjected to FEED A
(vital grower) while group B comprising of thirty
(30) chickens that were also subjected to FEED B
(top grower). Their weights were measured in kg at
the end of each week.
Scope and Limitations of the Study
This study is aimed at establishing the effects of
two types of poultry feeds on the weight of poultry
birds, using chicken as a case study. It involves an
experiment conducted for a period of eight weeks
where a sample size of sixty chickens was divided
3.
Assuming the profiles are coincident, are
they level? That is, are all means equal to the same
constant?
Or Equivalent: H03: µ11 = µ12 . . . = µ1p = µ21 = µ22 .
. . = µ2p . . . = µ2p = . . . = µg1 = µg2 . . . = µgp
If the two profiles are parallel it shows that the
mean of the feeds are the same considering all the
treatments applied together.
into two equal populations, each subjected to a
particular feed. Their responses were measured in
kilogram (kg) using weighing balance. The
variables x1, x2, x3, x4, x5, x6, x7, x8 stands for the
weights of chickens at the end of each week.
Review of Some Related Literature
There are several different multivariate test
statistics available for the test of parallel profile, all
of which will generally yield equivalent results.
Amongst the four commonly test statistics – namely
Wilks Lambda, Pillai’s Trace, Hotelling-Lawley
Trace and Roy’s Greatest Root; Wilks Lambda (λ )
is the most desirable because it can be converted
exactly to an F-statistics [18],[8] Johnson and
Wichern (2001) presented in their text a detailed
approach of this conversion and the exact
distribution of λ .
Bartlett (2001) in his work presented a
modification of λ for cases where the number of
groups is more than three (g > 3), as well as when
large sample sizes are involved. It is worthy to note
that (p – 1) would replace p.
Leboeur (1992) noted in his work – “Profile
analysis”, that experiment is conducted in a way of
.
observing two responses for a given population; the
same population is exposed to p-treatment at
successive times. And that we can formulate
successive times to enable us develop the question
of equality of means in a step wise procedure.
Hence, H01, µ1 = µ2 = … µn implies treatments
which have the same effect on the population;
hence, we test for coincident profile.
Profile analysis, according to Ott (1999), is a
specific style of Multivariate Analysis of Variance.
Tabacknick and Fidell (2006) stated that Profile
Analysis is equivalent to repeated measures,
because of its multiple responses taken into
sequence on the same subject(s).
Ohaegbulen E.U and Nwobi F.N [17] stated that
in poultry farming, the production of high quality
birds is always desired as this boosts the revenue of
the poultry farmer and showed how profile analysis
can be used to determine the feed with better
nutritive value to the poultry birds.
Croyle (2007) conducted a profile analysis on
self-harm experience among Hispanic and white
young adults. He compared the self-reported rates
of self-harm in 255 Non - Hispanic white (NHW)
and 187 Hispanic (predominantly Mexican
American). He observed that self-harm is relatively
common with about 31 % of the sample reporting
some history of self-harm. Rates and specific types
of self-harm did not significantly differ between the
Non-Hispanic and Hispanic groups.
Abdu, P.A., Mera, U.M. and Saidu, L. [1] had a
study on chicken mortality in Zaria and observed
that the use of profile analysis to conduct a chicken
mortality research is recommended.
Jensen and Dolberg [13],[6],[7] advocated for
using poultry as a tool in poverty alleviation. An
enabling environment must be established by
providing access to feed, vaccine, vaccinations
services, micro-finance, marketing and other inputs
and services. A village group, composed of
members of socially equal status, is an excellent
entity to disseminate improved technology, a costeffective entity to disseminate extension messages,
and a secure entity for disbursement of loans.
Rahman and Hossain, [19] showed that an
intervention with poultry production created a
relatively small decline in the overall poverty with
the proportion of extreme poor declining from 31 to
23% and the moderate poor stagnating around 29%.
Todd, [21] and Dolberg,[6] opined that poultry
activity is to be considered as a learning process for
the beneficiaries, but it has to be realised that one
activity alone is not sufficient to lift a family out of
poverty. The opportunities called as the enabling
environment must be available for the beneficiaries
to establish a small poultry enterprise, to minimize
the risks and to take up other income generating
activities.
Jensen [13] observed that about 70 % of the rural
landless women are directly or indirectly involved
in poultry rearing activities. He found that
homestead poultry rearing is economically viable.
Mack et al [16] opined that in order to increase
egg and poultry meat production there is a need for
increased investment guided by policies and
institutions that promote equitable, sustainable, and
environmentally friendly long-term outcomes as
backyard poultry make an important contribution to
poverty mitigation, it should be considered as any
strategy to improve rural livelihoods. Right policies
and investment, well designed and participative
development programmers can overcome the
constraints faced by the smallholder poultry
producers.
.
Karlan [14] opined that an enabling environment
would give all the villagers access to poultry farm
input supplies and services; pave the way for
disbursement of micro-credits in a cost-effective
way; facilitate easier formation of associations
through formalized village livestock groups; help
people acquire the skills that are required for a
business set-up.
Dolberg [7] reviewed poultry as a tool in poverty
alleviation focusing on experiences from
Bangladesh but survey and project work that has
been undertaken in India. Animal husbandry and
agricultural departments’ extension programmes are
hardly known or used by most poor people for
whom the poultry work is relevant.
Gondwe et al (2003) found that rural poultry is
raised and utilized by about 80 percent of the
human population, primarily situated in rural areas
and occupied by subsistence agriculture.
Research Methodology
Profile Analysis pertains to situations where a
number of treatments are administered to two or
more populations where all the responses must be
expressed in
Bujarbaruah and Gupta [3] reported that a flock
size ranging from 25-250 birds are reared across the
country under the village poultry system. They have
low production potential with only 40-80 eggs per
year but are less susceptible to most of the common
diseases requiring less veterinary care. In order to
meet the deficiency gap in poultry meat and egg
sectors, adequate and sustained efforts will have to
be made to improve the production efficiency of the
rural poultry which has been responsible to produce
40% of meat and 44% of egg requirement in the
country.
Krishna Rao [20] recorded that poultry are
inseparable from mankind and in the rural scenario
they do not need any land, are easy to manage,
regularly lay eggs, disease resistant and well
adapted to the harsh environment.
similar units and the responses for the different
populations are assumed to be independent of one
another. Suppose birds on diet one are observed
weekly for eight weeks, we can calculate the mean
using the formula below:
 X 11  ;


 X 12  Where the first subscript represents feed and the second subscript represents weeks. We plot the
 . 
X1 = 
 mean X 1 weights against the number of weeks.
 . 
 . 


 X 18 
.
Parallel Profile:
We assume that x 11 , x 12 ,..., x 1 n is a random
sample from Np(µ1, Σ ) and x 21 , x 22 ,..., x 2 n is
also a random sample from Np(µ2, Σ ).
C ( P −1) xP =
 µ 12 − µ 11 


 µ 13 − µ 12 


.
 , where Cµ = ( p − 1) * 1
C µ1 = 
1
.




.


µ − µ

1 p −1 
 1p
We can write Ho1 as
Ho1: Cµ1 = Cµ 2 Vs Ha1: Cµ1 ≠ Cµ 2 .
Instead
of
basing
our
test
on
the
observations x 11 , x 12 ,..., x 1 n , x 21 , x 22 ,..., x 2 n
we should use CXij; i = 1, 2 while j = 1, . . ., n.
To test Ho1; we calculate the Hotelling’s T2
[11].[12] as
(
)
−1
(

′  1 1 
T 2 = X A − XB C′ +  C SpooledC′ C X A − XB
n1 n2 

)
with
critical
region
T2
>
n + n2 − 2( p − 1)
t= 1
FP −1[n1 + n2 − p ]α , α = 0.05 , if we
n1 + n2 − p
reject Ho1, we stop and conclude that the profile are
not parallel and µ1 ≠ µ 2 , but if Ho1 is not rejected,
we test for coincident profiles given that the
profiles are parallel.
3.2
Coincident profiles.
If the profiles are parallel, they can only be
coincident if the total of element in µ1 is equal to
the total element in µ 2 .
.
The test statistic
(
−1
)
(

′ 1 1 
T2 =1′ XA −XB  +  1′(Spooled) I 1 XA −XB
n1 n2 

)
with critical region T > F1 , n1 + n2 − 2, α
If we reject Ho2, stop
2
3.3
Level Profiles
If the profiles are coincident the Xij, j = 1, 2,
…,n1 and X2j, j = 1, 2, …,n2 are sample space of
size n1 + n2 from Np(µ1, Σ ) where µ = µ1 = µ2.
The
test
statistic
is
given
as
−1
2
T = (n1 + n2 ) X ′C ′ (C S pooled C ) C X with critical
region
T2 >
3.4
(n1 + n2 − 2)( p − 1)
FP −1 , n1 + n2 − p, α
(n1 + n2 − p)
Mean and Pooled Covariance Matrix
n
The mean x =
∑x
i =1
i
n
Then for the respective group,
n
xA =
∑x
i =1
n
n
Ai
, xB =
∑x
i =1
Bi
Pooled =
A+ B
N1 + N 2 − 2
Analysis
A sample size of sixty chickens is involved in
this study. The chickens were divided into two
equal parts of thirty chickens each. The chickens
were labelled 1 to 30 for each of the groups. The
weights of each of the groups classified as A and B
are taken using a weighting balance for a period of
eight weeks labelled x1 x2, x3, x4, x5, x6, x7 and x8.
The weights over the weeks are shown in appendix
A.
The term profile is said to have been observed
by Hotelling (1936) to come from the practice in
applied works in which score on a test battery are
plotted in terms of graph or profile. Profile analysis
provides three types of information, level,
dispersion and the shape
Figure 1: A graph of sample means of Responses
per week
Sample profile for two types of poultry feeds on
the weights of poultry birds
Let A stand for feed A and B for feed B such that
the means X A and X B are the respective means
for the eight weeks under study.
n
The pooled venue is given by
4.2
Calculations For The Analysis.
N = 30, n1 = n2
Mean of Means for feed A and B is given as;
.
x =
=
The mean deviation of the two sample means are given by
(X
A
)
− XB =
While the sum of squares and cross products of each of groups are given by the symmetric matrix below
For feed A
A=
.
For feed B
B=
The Spooled (pooled covariance) matrices is given by
A+ B
Spooled =
where n1 = n2 = 30 such that n1 + n2 – 2 = 58
n1 + n2 − 2
Spooled approximated to 3 decimal places
I 1 SpooledI = [ Sum of elements in Spooled] = 0.667
(
)
(
)
I1 X A − X B = [Sum of elements in X A − X B ] = -1.64
.
(
)
C X A − XB =
=
Where c is a contrast matrix.
(X
)
1
A
X
− X B C1 =
1
−1 0 0 0 0 0 0 


 1 −1 0 0 0 0 0 
 0 1 −1 0 0 0 0 


 0 0 1 −1 0 0 0 
1
C = (0.66 0.85 1.05 1.27 1.41 1.55 2.17 2.26) 

 0 0 0 1 −1 0 0 
 0 0 0 0 1 −1 0 


 0 0 0 0 0 1 − 1
0 0 0 0 0 0 1


1
1
Therefore; X C = (0.19 0.2 0.22 0.14 0.14 0.62 0.45)
CSpooledC1=
.
| CSpooledC1| =
=
(CSpooled C1)-1 =
.
5.11 − 9.08 − 9.53 6.12
 74.57 3.77

 3.66 105.99 58.64 15.51 30.97 − 4.06
 5.04 58.55 105.37 48.62 58.00 15.10

=  − 9.02 17.55 49.76 84.04 67.35 9.64

 − 9.82 23.71 54.00 65.70 137.72 23.69
 6.07 − 5.57 14.26
9.50 22.97 98.38

 − 1.55 32.79 34.92 28.19 40.08 4.97
− 1.49 

33.54 
35.30 

28.94 

38.00 
4.46 

90.70 
(
)
(
Test For Parallel Profiles
Hypothesis:
Reject Ho1 : Cµ1 = Cµ 2 if T 2 cal > T 2 tab accept if otherwise.
(
Where T = X A − X B
2
)
′
−1
)
A ∼ N Cµ1 , CΣ1C 1 while B ∼ N Cµ 2 , CΣ 2 C 1
Where µ and Σ are the multivariate mean and
variance.
Since N is large, it is assumed that the two sampled
population are normal and the Hotelling’s T2
statistic can be used to carry out the various test,
similarly note that,
(
 1

1
C ′ +  C S pooled C ′ C X A − X B
 n1 n2 

)
T2
5.11 − 9.08 − 9.53 6.12
 74.57 3.77

 3.66 105.99 58.64 15.51 30.97 − 4.06
 5.04 58.55 105.37 48.62 58.00 15.10

 − 9.02 17.55 49.76 84.04 67.35 9.64
 − 9.82 23.71 54.00 65.70 137.72 23.69

 6.07 − 5.57 14.26
9.50 22.97 98.38

 − 1.55 32.79 34.92 28.19 40.08 4.97
− 1.49 

33.54 
35.30 

28.94 
38.00 
4.46 

90.70 
.
=
T2cal = 140.217
T 2 tab =
n1 + n2 − 2( p − 1) 2
T tab P−1[n1 + n2 − p ]α Let α be 0.05
n1 + n2 − p
T2tab = 17.709
Conclusion
where
Since T cal > T tab, we reject the hypothesis that
the profile are not parallel meaning that the mean
weight of chicken A is not the same with the mean
weight of chickens in group B considering all the
treatments applied together. This also implies that
feed A and feed B have different effect on the
chickens.

′  1 1 
T cal = I ′ X A − X B  +  I ′(S pooled) I  I X A − X B
n1 n2 

2
2
(
−1
)
(
)
and T 2 tab = T 2 tab 1 , n1 + n2 − 2, α
Let α = 0.05
Hence
Tests For Coincident Profile
Hypothesis
Reject Ho2 : I ′µ1 = I ′µ 2 if T
otherwise
 1

1 
T cal = (− 1.64 )  + 0.667
 30 30 

2
T tab. = 4.016
2
2
cal
>T
2
tab
accept if
2
−1
= 60.485757
Conclusion
We reject Ho2, since T2cal > T2tab and conclude
that there is no coincident profile. This means that
the response of chickens on feed A is not the same
with those of chickens on feed B.
where T 2 cal = (n1 + n2 ) X ′C ′ (C S pooled C ) C X
and
the
critical
region
(
n
+
n
−
2
)(
p
−
1
)
2
T 2 tab = 1
FP −1 , n1 + n2 − p, α
(n1 + n2 − p)
4.5
Test For Level Of Profile
Hypothesis:
Reject Ho3: Cµ = 0 if T2cal > T2tab accept if
otherwise
Let α = 0.05
Such that
is
.
T2
=
*
 74.57

 3.66
 5.04

 − 9.02

 − 9.82
 6.07

 −1.55
−1.49

33.54
35.30

28.94

23.71 54.00 65.70 137.72 23.69 38.00
− 5.57 14.26 9.50 22.97 98.38 4.46 

32.79 34.92 28.19 40.08 4.97 90.70
3.77 5.11
105.99 58.64
58.55 105.37
17.55 49.76
− 9.08
15.51
48.62
84.04
− 9.53
30.97
58.00
67.35
6.12
− 4.06
15.10
9.64
= 7364.4291
*
Hence T2cal =
while
the
critical
region
(n + n2 − 2)( p − 1)
T 2 tab = 1
FP −1 , n1 + n2 − p, α
(n1 + n2 − p)
is
Decision:
We reject the hypothesis that the
profile level since the T2cal > T2tab. The rejection of
the level profile hypothesis means that the chicken
on feed A and chicken on feed B do not have the
same level of response or that the average response
of the chicken to the respective feeds A and B are
not leveled.
Conclusion.
The analysis showed that the profile was not
paralleled and there is significant difference
between the two feeds A and B performance on the
weights of the chicken. The Average profile
of feed A was greater than that of feed B,
therefore, we select feed A as better than feed B.
Recommendation.
This method of analysis is recommended for
researchers trying to compare effects of an input on
the yielded results. It is also important for
researcher to note that time is of essence in this
form of research. This is to allow for proper
measurement of the weights as the chickens are
being feed with the respective feeds. This analysis
should be extended to more than two independent
populations.
T2tab = 17. 709
________________________________________
References
[1]
Adbu P. A and Mera U. M and Saidu L. (1992). A study on chicken mortality in
Zaria, Nigeria. In Proceedings. 19th World Poultry Congress, Amsterdam, Netherlands, 2, 151.
.
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
`
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
Bartlett, M. S. (2001). Further aspect of the theory of multiple regressions. Proc.
Cmb. Phil. Soc., 34: 33 – 34.
`Bujarbaruah K.M. and Gupta J.J. (2005). Family poultry for the development of
NEH region ICAR Research Complex Umiam, Barapani, Meghalaya IPSACON-2005
Croyles J. (2007 ). Some issues in the design of agricultural decision support
systems. Agricultural systems, 52(2/3): 355–381.
Del Ninno, C., Dorosh, P. A., Smith, L.C. and Roy, D.K (2001). The 1998
Floods in Bangladesh. Disaster Impacts, Household Coping Strategies, and Response. International Food Policy
Research Institute, Washington.
Dolberg, F. (2001). A livestock development approach for Rural Development:
Review of Household Poultry Production with Focus on Bangladesh and India Pro- Poor Livestock Policy
Initiative.
Dolberg Frands (2003). The Review of Household Poultry Production as a Tool
in Poverty Reduction with Focus on Bangladesh and India Pro- Poor Livestock Policy Initiative (PPLPI)Website:
Everitt, B. S. and Dunn G. (2001). Applied Multivariate Data Analysis. 2nd Edn,
Edward Arnold, London.
Gondwe Timothy N.P., Clemens B.A. Wollny, A.C.L. Safalaoh, F.C. Chilera and Mizeck G.G. Chagunda (2003)
Community-Based Promotion of Rural Poultry Diversity,
Management, Utilization and Research in Malawi.
Hotelling H. (1935) The Most Predictable Criterion” Journal of Educaitonal Psychology, 26, 139-142.
Hotelling H. (1936) “Relations between two sets of variables” Biometrika 28, page 321- 377.
Jensen, H. Askov (2000). Paradigm and Visions: Network for Poultry Production
in Developing Countries.
Karlan, D. S. (2002). Social Capital and Group Banking. Downloaded from
http://web.mit.edu/spencer/www/soccappaper.pdf
Leboeur, Y and Carlotte D.C., (2000). Body Weight and Low Back Pain Systematic Review of the Epidemiologic Literature.
Mack, D. Hoffmann and J. Otte (2000). The contribution of poultry to rural
development Organization of the United Nations, Viale delle Terme di Caracalla, 00100 Rome, Italy.
Ohaegbulen E.U and Nwobi F.N (2009) Poultry Feed brands solution using profile Analysis surreal of Applied Science
Ott, P., (1999). Multivariate Analysis of Variance: Profile analysis – an example
using SAS, biometric information pamphlet 60
Rahman, H. Z and Hossain. M (1995). Rethinking Rural Poverty: Bangladesh as a case study. UPL, Dhaka. Sage
publications New Delhi. Journal Article
Rao C. Krishna (2005). A rural poultry revolution for poverty alleviation in rural
India Former Animal Husbandry Commissioner, Govt. of India, Jaya Nagar , New Bowenpally, Secunderabad
Todd, H. (1999). Women in Agriculture and Modern Communication Technology.
Proceedings of a workshop, March 30-April 3, 1998, Tune Landboskole, Denmark.
Wald A. (1944) “On a Statistical Problem arising in the Classification of an Individual into
One of Two Groups” Annals of Mathematical Statistics 15 page 145-162
.
.
Information and Communication Technology (Ict) Integration Into Science, Technology,
Engineering And Mathematic (Stem) In Nigeria
A.A. Ojugo.+, A. Osika++., I.J.B. Iyawa* and R.O. Yerokun (Mrs.)**
+
Department of Mathematics/Computer Science, Federal University of Petroleum Resources Effurun,
Delta State (+234)8034072248, ojugo_arnold@yahoo.com, ojugoarnold@hotmail.com
++
Computer Science Education Department, Delta State College of Physical Education, Mosogar,
Delta State. +2348060328840 (osikaangela@yahoo.com)
**
Computer science Education Department, Federal College of Education (Technical), Asaba,
Delta State. iyawaben@hotmail.com, agapenexus@hotmail.co.uk
Abstract
As Nigeria aspires for technological growth, positive changes need be made by placing proper educational values
towards Science, Technology, Engineering and Mathematics (STEM) education. Some problems faced by STEM include
lack of qualified teachers, curriculum, the misconception that STEM education is reserved for the less intelligent in the
society, amongst others. Need thus arises, to develop alternative strategies to alleviate such problems. Bridging ICT and
constructivism will exponentially change educational processes of both teachers and students to succeed as well as
present new forms of learning environment. It will also unveil the power shift in educational structures; equip students to
become knowledge producers rather than consumers. Thus, an ICT-Constructivist rich class will help teachers
enfranchise, emancipate students academically with a framework that measures quality of engaged student’s learning.
This paper aims to reveal links between effective use of ICT and the long neglected theory of constructivism in the area of
STEM Education.
Keywords: Constructivism, Rationale, informatics, productivity, meida literacy, lifelong.
___________________________________________________________________________________________
1.0 Introduction
STEM has now become an integral part of our
culture – in that actualizing our current societal
goals and those of our generations ahead will be a
mirage unless we have excellent understanding of
STEM. [1] points that STEM has become our
heritage and mankind’s hope. Thus, the mission of
today’s education must ensure students are
scientifically literate to cope with technological
changes of the data age. [2] notes STEM goals as:
(a) provides preparation for training in science and
mathematics, (b) provides basic mathematics and
science literacy for everyday living, (c) provide
basic skills, attitude to prepare us for technological
developments and (d) help stimulate and enhances
creativity.
Education is the art of transferring knowledge
from a teacher to learner within a physical
environment (called school with classroom) and
.
interrelationship that [3] calls a system of factors,
which jointly affect learning individuals of cultural
differences. The classroom provides the structure in
which learning is organized and the school has three
major components namely: learner, teacher and
administrator. There are basically two styles of
education delivery namely Traditional (a teacher
employs face-to-face, oral method in which the
teachers pass knowledge to a learner), and
Alternative Delivery (learners can construct new
knowledge and meaning from previous concepts and
beliefs with methods and strategies that involves
media literacy. The latter is more concerned with
what a learner does and understands rather than
teacher’s input. Thus, the use of equipments
becomes focus of study (technology education) and
educational support (educational technology) as in
figure 1 [2,4].
Computer System
Hardware/Software
Non-Interactive
Technology
Students
Teachers
Traditional
Classroom
Physical Features
of a Classroom
Curriculum
Figure 1 shows a constructivist-class model
[5] laments that STEM education is not given serious
attention as it is misunderstood by educators and
stakeholders in Nigeria, who believe that STEM
education are for those who cannot pursue academic
programmes. Thus, proper values must be placed on its
need to help attain the desired growth as today’s
industrialized nations employed the skills of both the
educated and less educated in their growth toward ICT.
Technological advancement in Nigeria today, is a sad
reflection of the poor quality of STEM education that
still receives stigmatization in our educational system,
hindering our expected technological progress.
1.1. Objective of Study
The study objective is: (1) seek alternative exemplary
educational practices, (2) showcase outcome that
describes individual/group adoption patterns of
technology in learning as well as (3) show evidences
linking technology integration, quality students learning
and staff development.
1.2. Statement of Problem
This study investigates learning in technology
rich class (constructivist) versus the traditional class
(non-constructivist) via creaction of a constructivist
and non-constructivist groups. Will constructivism
make a difference via meaningful, engaged-learning
and transferable knowledge by students, in their
own context? It will also seek the extent of the
success in constructivist learning as effective
method to boost student’s learning, performance
and achievement.
1.3. Research Question
The study aims to determine:
a. Extent of student’s achievement and attitude in a
constructivist and non-constructivist groups as reflected
in their learning of STEM.
b. Impact of ICT on various learning processes.
c. What is the rationale behind ICT integration.
2. 0 ICT Framework In Education
The provision of modern technological equipment at
all school levels varies due to their various levels of
preparedness – both by teachers and students. A look at
students’ ability in obtaining the necessary flexibility in
the world of information closely correlates amongst
others, level of information setting of schools as this will
help provide schools with various expanse of data in
.
printed and electronic format via regular updates and
networks as well as those directed by experts in the
informatics field.
ICT integration into education results in great reforms
to the learning process; and educators who advocate
such reforms, opines that such learning is informed by
constructivism [6] which pleads the need for students to
develop high thinking skill and the failure of the current
schooling methods to provide such opportunities [7-8].
Thus, a critical factor to ICT integration is
constructivism, so that learning takes place as the
learner completes tasks for which media support is
required and used to maintain such learning environment
and learners [9].
Technology creates ideal learning. Thus, [10] notes
it has been ignored or its past implementation has failed
widely – as it creates a learner-centered, learning
environment with a belief that they learn more from
what they do or think rather than the teacher’s input. But
we must take care not to allow the dynamic nature of
technology overshadow the enduring nature of learning
and or the ever-increasing knowledge base about
learning [7, 4].
[11] notes the concept of constructivism as one in
which a learner has the ability to actively construct
knowledge as he learns. It also emphasizes knowledge
as a construction of reality in the learner’s mind because
knowledge is a dynamic adaptation towards an
interpretation of experience. It supports many
interpretations to reality based on knowledge
constructed from experience and media-rich class. It
focuses on knowledge construction rather than
consumption – as the learner constructs data from
experiences and mental beliefs, interpreting events
accomplished outside the mind. We thus see the world
we describe rather than describing the world we see.
When integrated across curriculum, it provides
appropriate level of difficulty due to its tasks that are of
real world relevance with engaged-learning and teachers
becoming knowledge guides [12] as seen in figure 2.
2.1. Constructivism
Today’s education is transformed by new
technologies that provides large amount of data to
learners, coupled with the fact that knowledge doubles at
increasing speed requiring a transformative method to
education. Its challenge is that educators and learners are
suspicious of the educational practices as it differs from
what they are used – as constructivism removes
statewide assessment, which traditional model promotes
by aligning tests. The issues of fund shortage, unclear
vision to keep the change from occurring rapidly as well
as teachers not having a good understanding of how ICT
works as they are charged with the duty of emancipating
students, are in continual resolution. This paradigm shift
will require teachers’ retraining, role redefinition as well
as acculturation to put this systemic change in place –
even though it be slow [13].
[14] notes some of the known principles of the
constructivist learning as thus:
1. Learning is active and consists of meaning
construction and constructing system for meanings.
2. Knowledge construction is mental – since it
happens in the mind of the learner.
3. Language influences greatly what we learn.
4. Learning is a social activity associated with
connections the learner makes with others.
5. Learning is contextual – as learner cannot learn
isolated facts or theories in abstract ethereal land,
separate from real world situations.
6. Motivation is a key in learning to help us know how
the acquired knowledge can be put to use.
7. Knowledge is important – as we cannot comprehend
new ideas without some structure developed from prior
knowledge to build on.
8. Learning is not instantaneous but takes time – as
the learners must revisit principles, reflect on them and
using them as often as possible.
2.2. New Paradigms: A Constructivism Class
Education, transformed by new technologies yields
the following paradigms, when adopted:
.
a. Shift from whole class to smaller groups
b. Teachers coach weaker students rather than focus on
all as with traditional settings. Coaching occurs rather
than lecture and recitation.
c. Students become actively engaged, cooperative and
less competitive
d. Students learn differently than simultaneously.
e. Integrate visual/verbal thinking rather than primacy
of verbal thinking as in traditional class.
Thus, educators, parents and learners will become
suspicious of the educational practices as it differs from
what they are used. This is becasue such constructivist
learning removes statewide, aligned assessment – unlike
traditional model. Thus, learners will take standardized
tests, which does not assess what they are learning but
what new meaning they derive of concept. Class
structure will become more fragmented and problems
will abound due to lack of funds and unclear vision to
keep this systematic change from occurring as rapid as
possible. Teachers charged with the duty of
emancipating students, may not have adequate
understanding of how these technologies work and the
amount of data available as such paradigm shift requires
staff retraining, their roles re-defined to inform them to
think about why they do what they do as well as funding
[15-20].
2.3. Challenges of Constructivism
The common challenges of constructivism as:
1.
Nativism: Cultural constructivism promotes
nativism and language, which primarily distorts the
fundamental unity in education generally. It thus denotes
knowledge as meanings conveyed by learners in
different tongues – though referring to same state.
Meanings, applied are inseperable of linguistics; though,
science views meaning as an objective states that
transcends such linguistic boundaries. It thus proposes to
recreate nature to suit cultural and linguistic boundaries.
But, the nativism and empiricism of science are too
parallel and may never meet at internationalization and
globalization [21].
2.
Knowledge Territorialism: A concept of false
belief that Africans who live enclosed cannot yield
scientific discourse. Knowledge transcends the idea of
cultural boundaries – such that ideas from varied
ccultural perspectives must converge a conventional
consensus due to similarities over their differences –
though, such similarities by virtue of their many
appearances, imprints itself upon the mind; while
individual differences that changes between cases, fails.
Faraday, Newton amongst other scientific inventors
made their ideas to transcend ethnic boundary. Thus, we
must deterritorialize our seminars to hold global focus;
instead of its localized viewpoint [5].
7.1
Globalization: Poor globalization in
STEM is often misconstrued as nativism. Scientist
must ensure unity via interaction and exchange of
concepts, innovations and skills among experts world
over. This will urge individuals and research
organizations not to be localized and restricted by
culture. Competition must be encouraged, with
knowledge circulation a rule. Thus, cultural
constructivism in education is faced with the challenge
of capacity building and establishment of research
networks with Africans in Diasporas and with other
worlds [21].
3. 0 Method And Materials
This will be discussed under the following heading:
3.1 and 3.2 respectively.
3.1. Researchable Model:
The researchers will adopt [22] researchable model as
redesigned by [5] for STEM education. Teaching
method will be divided into constructivist (via
laboratory, Internet to allow online interaction) and nonconstructivist (normal classroom) groups. Curriculum
content in focus: Mathematics (Geometry), Biology
(Reproduction), Physics (Pendulum) and Chemistry
(Titration).
Feats
to
be
measured
include
teachers’/students’ attitude and involvement.
.
groups (i.e. 30 students and 30 teachers) – and the
sstratified sampling method adopted, in the selection
criteria to alleviate falsehood of results obtained as well
as give a fair representation of population.
3.2. Population
Nigeria, is today divided into 6 geo-political zones:
South East (SE), South-South (SS), South-West (SW),
North-North (NN), North-East (NE) and North-West
(NW). The scope of the study is limited to tertiary
institutions in the 6 geo-political regions. Two schools
were chosen from each geo-political zones: Federal
University of Petroleum Resources Effurun and
University of Port-Harcourt (SS zone), University of
Nigeria Nsukka and Imo State University Owerri (SE
zone), Osun State University and University of Ibadan
(SW zone), University of Agriculture Maiduguri and
University of Jos (NW zone), Ahmadu Bello University
Zaria and Nigerian Defence Academy Kaduna (NN
zone) and lastly, Idah Polytechnic Idah and University
of Illorin (NE zone). From these, 15 samples each were
selected for the constructivist and non-constructivist
4. 0 Data Analysis And Findings
Pedagogical practice and extent of involvement will
determine if a group exhibits characteristics of a
constructivist class or not. Its outcome is measured via
assessment of student’s performance as seen in figure 3.
For analysis, mean (X) and standard dev. (SD) are used
with results discussed below in:
4.1. Research Question 1: ICT integration impact on
students and what extent their achievement is
reflected in their attitude?
Table 1a. Mean score of student acheivement in two gruops
Pre-Test
Groups
Experimental/Construc
tivist
Control/No
n constructivist
Post-Test
No.
Mean
SD
Mean
SD
720
13.84
64
3.1
24
21.43
33
4.1
50
720
10.62
5
2.5
61
19.40
27
3.4
84
Table 1b: Student’s attitude towards STEM in two groups
Groups
No.
Male
Female
Experimental
Constructivist
Control
or
Constructivist
or
720
10.1663
9.9823
720
11.0001
10.0011
Non-
Table 1a shows the differences in students’mean score
achievement. At post-test, students in the constructivist
group had a higher mean scores than their counterpart in
non-constructivist group. This is attributed to the
exposure they had in the use of ICT in learning. While
table 1b shows a significant difference as male students
in the non-constructivist group exhibited higher attitude
to STEM than their counterpart in constructivist group.
.
This can be a result of technophobia as they were
introduced to a completely, new-pedagogical learning
pattern and scenario. Thus, ICT supports learning with
technology literacy, high academic emancipation from
teachers, increased motivation for learning, improves
their achievement in core subjects as measured by tests,
increases their engaged learning and interdependence
that allows them develop skills that are associated with
time and resource management, concentration, selfdiscipline, attention to defined task and ability to follow
instructions. Any change in role and requirement for
new sets of skills to be introduced/supported, must be
carefully done with consideration for learners with
opposing perception and poor past experiences. It is also
discovered that students use ICT in different ways
because more access requires greater personal
responsibility that is lacking in some students.
teachers and attitude reflection?
Table 2. Teachers attitude towards STEM in two groups
Table 2
Groups
Experimental
Constructivist
Control
or
Constructivist
or
Non-
No.
Male
Female
720
14.663
9.660
720
11.001
8.912
Table 2 shows significant difference in the experimental
group as teachers are more at home with the idea that
such courses must and should be taught in technologyrich classes. ICT integration requires a complex change,
as teachers must remain instructional leaders to aid
human interaction, motivation and to retain their
established influence over directing class activities.
Teachers must reflect on ICT’s impact on their roles and
on the learner. Teachers seeking to employ ICT note the
issue of curriculum, learning materials, ICT usage in
classroom, student roles and behaviour – as they must be
provided with information access that leads to increased
interest in teaching and experimentation. Such learning
requires more collaboration from administrators, parents
and students with proper planning, energy, skills
development and knowledge of ICT. This will lead to
greater productivity via more engaged time for learners
and presents a pedagogy with strategies that are: (1)
learner-centered and active, (2) more cooperative and
collaborative, (3) learning based on great information
access and source, and (4) create in learners, the need
for interdependence.
4.3. Research Question 3: ICT impact on Learning and
Its Environment.
ICT offers new learning methods, allow teachers to rely
on educational theory and past knowledge of educational
situation to aid better decisions about what/how the
learning environment will look like as well as improve
overall effectiveness of learning environment. Its
positive impacts are: (a) class-size reduction, (b)
increased learning time, (c) better performance via the
use of cost effective computer aided instruction (CAI)
programs, and (d) significant gains in learning as ICT
learning is mediated via components like curriculum,
teachers and pedagogy. ICT allows proper investigation
of real world applications with vast amount of data
access, and tools to analyze and interpret data as well as
broaden and deepen knowledge. It also allows active
participation and proper assessment of class activities.
Students’ engagement with curriculum will increase and
afford them opportunities to create their own data and
represent their own ideas. Simulation programs will help
provide learners with learning experiences as they
interact offline (with computers) or online (with others).
Thus in all cases, students has more influence on
learning as activities becomes more responsive to
learners’ need to better facilitate development of
theoretical framework and assist in deeper levels of
learning.
school Curriculum.
.
Curriculum is a two-way rrelationship – in that ICT is
used to cconvey curriculum and vice versa, change the
contents of curriculum. Report shows that effective use
of ICT to support learning is a function of the
curriculum content and instructional strategy such that
when appropriate, contents are addressed using
appropriate strategies that students and teachers will
benefit from. The impact of ICT to curriculum can be
viewed in terms of (a) declarative knowledge that
describes events by specifying properties that
characterizes them, and (b) procedural knowledge that
focuses on the processes needed to obtain a result or
knowing how. This model is dynamic with interactive
multimedia that provides visualization and analytic tools
that are currently, changing the nature and inquiry into
STEM. These changes affect the kind of phenomena
considered and the nature of argumentation and
acceptance of evidence. Thus, curriculum must remain
relevant to societal needs so that while at school, it
forms the learner’s foundation that helps them connect
in meaningful and motivating ways as they apply it to
their workplace – since at some stage, it becomes a
trajectory needs to connect them with non-school
discourses.
the Education System.
Schools must provide infrastructure and support for
learning to help maintain ICT integration and
constructivist learning. [17] highlights seven
requirements for ICT as: (1) Suiting technology to
educational goals and standards, (2) vision for
technology use to support curriculum, (3) provide inservice and pre-service training, (4) provide teachers
time to plan and learn how to integrate the technologies,
(5) ensure access to the appropriate technology, (6)
provide administrative support for technology use, and
(7) provide technical support for technology use. These
falls into 5 areas as noted by [5]: (a) Provision of
infrastructure, hardware and software, (b) Schooling, (c)
Curriculum, (d) School design, organisation, policies
and practices, and (e) Ttechnical support for teachers.
4.6. Discussion Of Findings
The study results show that there is significant
difference exhibited by mean scores of students and of
teachers in constructivist group over their counterparts
in non-constructivist group. This confirms the view [5],
that learning via the constructivist model paves way for
meaningful, engaged learning and active participation
and serves as motivational factor in learning.
4.7. Rationale For ICT Integration
The rationale is whether in practice, it has positive
impact and must lead to a system that decides what
students, teachers and the school, aims to achieve.
[23-24] notes that 3 main rationales are thus:
1. Education Productivity: Ratio of output over input
is the quantity/quality of learning demonstrated by the
student over cost. With the proper selection of input,
learning is optimized to increased outcome. Productivity
cannot be based on the fact that ICT media are
expensive to install – as cases may arise in where
ttechnology becomes the solution to a problem. If part of
the curriculum is not completed for lack of technology,
its associated outcome and productivity becomes zero.
2.
Technological Literacy – ICT helps to address
problems in curriculum. Education technology is
selected on the basis that it has best feats for
implementing the curriculum, as there is always a twoway relationship between curriculum and educational
technology. Firstly, policy makers decide what to learn
(curriculum), after which technology and the method to
be used is determined by the intended curriculum.
Secondly, new technologyy adds new feats to
curriculum contents; while making some contents
obsolete.
3.
Student’s Learning Support – There are much
potential for the use of computers in learning but
whatever the rationale, requires a critical evaluation on
.
the part of students. We must bear in mind these criteria
to be met on the learner’s part: (a) managing high
quality educational programmers requires large amount
of data, which teachers must effectively help students
manage, (b) Access to resource materials linked to
teaching and learning (online and offline), and (c)
computer literacy.
5. 0 Summary And Conclusion
This study contributes in four ways: (1) its outcome
gives a description of individual and group adoption of
technology for teaching and learning across the various
educational levels, (2) images exemplary practices for
teaching, learning and research; and (3) it links ICT
integration, engaged students learning as well as staff
professional development. Teachers and aadministrators
having gained insight, must rise and equip themselves to
make decision that will avert these problems, as they
hold the keys to students success in this new education
plan.
Its significance is both theoretical and practical as it
highlights the folowing:
a. Increases awareness/application of ICT theories –
naming issue and cchallenges with widespread adoption
of ICT in education and outcome of curricular across
subject areas in schools.
b. Hhighlights an adoption mode documentation and
feats of administrators/teachers willing to integrate ICT
with the support of network facilities. Our results shows,
use of shared data instrument for widespread access by
both teachers and students, based on engaged learning
and the stages of technology adoption, will form the
basis for the next step in the planning and
implementation processes at each school. A useful
information to all stakeholders in educations.
c. Education mode will provides the efficacious,
much-needed images of integration for engaged learning
– as such knowledge is useful, at organizational and
individual level for staff development and ffurther
research in such areas.
6. 0 Recommendation
a. Government must fund the integration and provide
infrastructural support – as reform is not just the
provision of ICT equipments.
b. Curricular reforms must be made to reflect ICT
integration with reviews presented to the government for
proper assessment and /implementation.
c. Staff development schemes organized by school
administrators to equip and redirect teachers’ focus
to emancipatee students. This scheme and
retraining, will aid teachers to better understand
their new and expected role as well as will help
them navigate a fully ICT integrated curricular.
References
[1]
[2]
H.A. Rowe., “Learning with personal computers,” 1993, Hawthorn: Australian Council for Educational
Research.
A.A. Ojugo., F.O. Aghware and E.O. Okonta, “Re-engineering Nigerian education via ICT” EBSU Sci.
Edu Jour., ISSN: 0189-9015, 2005, pp. 15-21.
.
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
G. Salomon., “Differences in pattern: studying in computer enhanced learning environments,” 1994,
Heidelberg, Springer Verlag.
A.A. Ojugo, F.O. Aghware and A.O. Eboka., “Role of ICT and CAI in Nigerian education,” 2006,
IRDI, ISSN: 1597-8028, Vol. 2, 25-31.
A.A. Ojugo, F.O Aghware., E.O. Okonta., R.O Igbunu and C. Onochie., “Constructivism: Aid to
teaching/learning of computer science in Delta State, Nigeria,” 2008, ASSEQEN, ISSN: 0331-4790, Vol. 5(2),
pp. 122-127
R.W. Clouse and H.E. Nelson, “School reforms, constructed learning and educational technology,
2000, J. Edu. Tech. Sys., Vol 28 (4), 2000, pp. 289 – 303.
J. Campione, A.L. Brown and M. Jay, “Computers in the community of learners, 1990, Springer-Verlag,
Berlin.
D. Loader and L. Nevile., “Educational computing: resourcing the future,” 1991, Incorp. Asso. of
registered teachers of Victoria, VC.
L.S. Vygotsky., “Mind in society, development of higher psychological processes, 1978, Harvard press,
Cambridge.
C. Glickman. “Pretending not to know what we know, 1991, J. Edu. Leadership, Vol. 48(8), pp. 4–10.
J. Dewey, “Democracy and education,” 1966, Free press, New York.
R.H. Strommen and G.A. Lincoln. “Role of constructivism in the changing state of education,
application at Montgomery school for children. 1992, J. Edu. Tech, Vol. 32, 1 –18.
E. Vosniadou, “Implementing new education technologies, 2004, Princeton, New Jersey.
BECTA, “Impact of Information and communication technologies on pupil learning and attainment,”
2002, ICT in school research and evaluation series – No. 7, DfES.
Committee on Development of Science Learning (Ed.), “How people learn: Brain, mind, experience
and School. 2000, National Academy press, Washington DC.
E. Decortes, “Learning with new information technologies in schools: perspectives from the psychology
of learning and instruction,” 1990, J. Comp. Aided Instru., Vol 6, pg 69 – 87.
A.L. Brown, “Advancement of learning,” 1994, J. Edu. Researcher, Vol. 23(4), pp. 4 – 12.
B. Collis, “Using information technology to create a new educational situation,” 1989, UNESCO
Congress on Edu. and Info., pp. 19.
T. Laferriere and R. Bracewell., “Benefits of using ICT for teaching and learning in K-12/13
classrooms,” 1999, SchoolNet Canada [online] www.schoolnet.ca/snab/e/reports/research.asp.
C. Lankshear and I. Snyder., “Teachers and technoliteracy,” 2000, Allen and Unwin, St. Leonards,
NSW.
O. Abonyi, “Cultural constructivism in science education: issues of 21st century In Nigeria,” 2005,
NERDC: Lagos, Vol.12(2), p. 172 – 183.
R.A. Yager., “Constructivism: an instructional strategy to reposition teaching and learning of
mathematics in secondary schools,” 1991, New York: Free press.
P.C. Newhouse., “Impact of ICT on learning and teaching,” 2006, McGraw Hill publication, New York.
Y.U. Ilo, “New media for education and training”, 2004, Ferlam publishers, Geneva, 31-38.
.
acquires
Student
To live in
Learning
Community
responsible to
demonstrated through
mandates
Educational
element of
Learning
Outcome
Curriculum
Pedagogy
provides
resources to
Content
Supports
use to
deliver
School
element of
support
influence
provides
determine
Learning Environment
• Physical
• Psycho-social
• Learning community
Teacher
element
of
element
of
ICT Resources
• Software
• Hardware
has capabilities
with
has capabilities
with
supports
Figure 2 shows the concept map indicating the relationship between the learning environment entities and external entities
Methods
Constructivist
Learning
Environment
Curriculum
Content and
Coverage
Classroom
STEM Concept
NonConstructivist
Laboratory and
Association
Participant’s
Behaviour
Student
Attitude
Student
Involvement
Learning
Outcome
Science,
Tech., Eng.,
and Maths
(STEM)
achievement
test in
schools
Teacher
Attitude
Teacher
Involvement
Figure 3 shows a researchable version of constructivist learning models using Yager’s design
.
Comparative Analysis of the Functions 2n, n! and nn
Ogheneovo, E. E.; Ejiofor, C. and Asagba, P. O.
Department of Computer Science, University of Port Harcourt, Port Harcourt, Nigeria.
edward_ogheneovo@yahoo.com, ejioforifeanyi@yahoo.com, pasagba@yahoo.com
Abstract
In this paper, we have attempted to do comparative analysis of the following functions: 2n, n! and nn. We
analyzed these functions, discussed how the functions can be computed and also studied how their
computational time can be derived. The paper also discussed how to evaluate a given algorithm and determine
its time complexity and memory complexity using graphical representation of the various functions, displaying
how the function behaves graphically. However, it was noticed that when data are inputted into these functions,
they gave cumbersome outputs that make it impossible to determine the execution (computational) time for the
functions. We plotted a graph by taking a snapshot of the integer values n = 1 to 10 to compute the functions of
2n, n!, nn. From the graph, we noticed that 2n function had lower growth value; nn had the largest growth value
and n! had slightly greater increase in growth than the 2n function. From our result, the execution time cannot
be computed due to the largeness of the outputs. However, we were able to determine the function with the
highest computing time and discovered that the time growth for the functions differs from one to the other.
Keywords Algorithm, Pseudo code, Exponential functions, Recursion, Complexity.
_________________________________________________________________________________
or Y = f(X)
1.0
Introduction
Functions pervade all areas of mathematics and
its applications. A function is a relation which
associates any given number with another number
[5]. Functions can be defined in several ways. We
define a function from the set X into the set Y as a
set of ordered pairs(x, y) where x is an element of x
and y is an element of Y such that for X in x there
is only one ordered pair (X, Y) in the function P.
the notation used is
f : X → Y or Y = f(x)Y or X → f(x)
A function is a mapping or transformation of x
into y or f(x). The variable x represents elements of
the domain and is called the independent variable.
The variable y representing elements of the range
and is called the dependent variable (Clarke, 1996).
The function y = f(x) is often called single valued
function since there is a unique y in the range for
each specified x. the converse may not necessarily
be true, y = f(x) is the image of x.
Often, a function depends on several independent
variables. If there are n independent variables x1, x2,
.
x3, …, xn and the range is the set of all possible
values of corresponding to the domain of (x1, x2, x3,
…, xn). We say that y is a function of xi’s, y = f(x1,
x2, x3, …, xn). Letters other than f may be used to
represent a function [3]
2.0
Exponential Functions (2n And Nn)
Exponential functions are perhaps the most
important class of functions in mathematics. We
use this type of function to calculate interest on
investments, growth and decline rates of
populations, forensic investigations as well as in
many other applications (Constatinescu, 2004).
The application of this function to a value x is
written as exp(x). Equivalently, this can be written
in the form of ex, where e is a mathematical
constant, the base of the natural logarithm, which
equals approximately 2.718281828, and is also
known as Euler’s number (Schmidt and Makalic,
2009).
As a function of the real variable x,
the graph of y=ex is
always
positive (above
the x axis) and increasing (viewed left-to-right). It
never touches the x axis, although it gets arbitrarily
close
to
it
(thus,
the x axis
is
a
horizontal asymptote to the graph). It’s an inverse
function [2].
Exponential growth is "bigger" and "faster" than
polynomial growth. This means that, no matter
what the degree is on a given polynomial, a given
exponential function will eventually be bigger than
the polynomial. Even though the exponential
function may start out really, really small, it will
eventually overtake the growth of the polynomial,
since it doubles all the time [1]
2.1
Factorial
The number of sequences that can exist with a set
of items, derived by multiplying the number of
items by the next lowest number until 1 is reached.
In Mathematics, product of all whole numbers up to
0 is considered. The special case zero factorial is
defined to have value 0! = 1, consistent with the
combinatorial interpretation of their being exactly
on way to arrange zero objects. The factorial of all
non-negative integers less than or equal to n.
n! = n(n-1)(n-2) … 3 x 2 x 1.
where n! represents n factorial
n = number of sets (items)
For instance, the factorial operation is encountered
in many different areas of mathematics, notably in
combinatory, algebra, and mathematical analysis
[13]. Its most basic occurrence is the fact that the
definition of the factorial function can also be
extended to non-integer arguments, while retrieving
its most important properties [4].
3.0 Computing Times Of Some Growing
Functions.
The time for different functions differs from one
to the other. Some functions have a greater time
growth than others. For example, we consider the
figures 6 and 7 (the graphs) below; it shows how
the computing times for 6 of the typical functions
on the table grow with a constant equal to 1. You
will notice how the times 0(n) and 0(nlogn) grow
much more slowly than the others[9]. For large
data set, algorithms with a complexity greater than
0(nlogn) are often impractical [14], [8].
An algorithm which is exponential will only be
practical for very small values of n and even if we
.
decrease the leading constant, say by a factor of 2
or 3, we will not improve the amount of data we
can handle significantly [7].
To see more precisely why a change in the constant,
rather than to the order of an algorithm produces
very little improvement in running time, we will
consider the figure below:
Fig 6: Graphical representation of the functions 2n,n3, n2.
Fig 7: Graphical Representation of the functions nlog2n, n, log2n.
Comparing the Growth of Functions 2n,
N!, Nn
Due to the fact that the execution time of
function 2n, n!, nn is unreliable, and even though we
had to give extra computing load to these functions,
we still could not have a visible execution time.
However, we decided to compare the growth of the
functions in terms of the magnitude of the values
they compute.
We implemented these algorithms by using a
program in the form of Turbo C++ program.
When we entered the consecutive values for n
from 1 – 150, the program generated growing
output values for the various functions. We noticed
that the program could not generate an output for
the function when n is greater than 150. We decided
to change the type of the value returned by the type
of the value assigned to the local variable temp. The
program was rerun and we noticed that although it
3.1
generated values for n > 200, there were some
errors (problems) with the results of some of the
functions. We noticed that the result generated by
n! and nn started to generate negative integer values
from n >= 20. In addition, we also discovered that
after some time, n! started generating 0 as output.
In other words, it stopped generating results as we
continued increasing the integer values for n.
4.0
Discussion Of Results
In this section, we are going to make a certain
assertion about the behaviours for the growing
functions of 2n, n!, nn and we also use a graph
plotted of the functions against the values of n to
discuss our findings.
•
With the graph of the growing function of
2n, n!, nn depicted in figure 8
Figure 8: Graphical representation of 2n, n!, nn
We plotted a graph by taking a snapshot of the
integer values n= 1 to 10 to compute the growing
functions of 2n, n!, nn. In this graph, we discovered
that the 2n function had a lower growth of value
than the n! and nn functions. We also noticed that
the nn had the largest growth of values than the
functions 2n and n! We observed also that n! had a
slightly greater increase in growth than the 2n
function.
5.0 Conclusion
The execution time of functions cannot be
calculated due to the largeness of the outputs when
a value is inputted. However, we were able to
determine the function with the highest computing
time from the altitude of the curves in the graphs
plotted. The time growth for functions differs from
one to the other. Some grow much slowly than
others while others are immensely fast. However,
14
.
the execution time could not be computed for the
functions 2n, n!, and nn.
______________________________________________________________________________
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
Abramowitz and Stegun, (1972), Exponential Functions, In Handbook of Mathematical Functions with
Formulas, Graphs, and Mathematical Tables, New York, Dover, pp. 69 – 71.
Ahlfors, L. V. (1953), Complex Analysis, McGraw-Hill book Company Inc., U.S.A., pp. 56 – 80.
Anyanwu, S. A. C. (2002), Elementary Modern Algebra and trigonometry, Markowitz, Centre for
Research and Development, Port Harcourt, pp. 203.
Borwein, P. (1985), Complexity of Calculating Factorials, Journal of Algorithm, Vol. 6, pp. 376 – 380.
Clark, G. D. (1996), Basic Calculus and co-ordinate Geometry for First Year University Students
, GODSONS Books, Port Harcourt, pp. 1 – 3.
Constantinescu, E. (2004), Inequalities for Logarithmic and Exponential Functions, General
Mathematics, Vol. 12, No. 2, pp. 47 – 52.
Gerety, C. and Cull, P. (1986), Time Complexity of the Towers of Hanoi Problem, ACM
SIGACT News, Vol. 18, No. 1, pp. 80 – 87.
Heileman, G. L. (1996), Data Structures, Algorithms and Object-Oriented Programming, MoHill Book Co., Singapore, pp. 23 – 46.
Horowitz, E. and Sahni, S. (1978), Fundamentals of Computer Algorithms, Library of Congress
Cataloguing, pp. 20 – 39.
Kruse, R. C. (1994), Data Structures and Program Design, Prentice-Hall, New Jersey, pp. 34 –
56.
Sahni, S. (1998), Data Structure, Algorithms and Application in C++, Mc-Hill Book Co.,
Singapore, pp. 15 – 39.
Schmids, D. F. and Makalic, E. (2009), Universal Models for the Exponential Distribution,
IEEE Transactions on Information Theory, Vol. 55, No. 7, pp. 3087 – 3090.
Wikipedia, the Free Encyclopaedia, Factorials.
Wirth, N. (1976), Algorithms and Data structures, prentice-Hall, New Jersey, pp. 20 – 47.
.
Appendix A: Program Codes
#include <iostream.h>
#include <math.h>
#include <time.h>
#include <stdio.h>
#define size 1000
double factorial(long);
int main()
{
long number, fact;
double expon[size], factn[size], npowern[size];
cout<<" \n Enter the value of n: ";
cin>>number;
if(number < 0 )
{
cout<<" You have entered a wrong input!"<<"\n";
cout<<"\n Program stops!";
return 0;
}
for(int i = 1; i <= number; i++)
{
expon[i] = pow(2, i);
factn[i]=factorial(i);
npowern[i]= pow(i, i);
}
cout<<"\tn 2 ^ n
n!
n ^ n \n";
cout<<"\t=== ===== === ===== \n";
for(int k = 1; k <= number; k++)
cout<<"\t"<<k<<" "<<expon[k]<<"
"
<<factn[k]<<"
"<< npowern[k]<<"\n";
getchar();
return 0;
}
double factorial(long n)
{
double temp; if(n == 1)return 1; if(n > 1)temp=n * factorial(n - 1); return temp; }
.
Implementation of A Collaborative E-Learning Environment On A Linux Thin-Client
System
Onyejegbu L. N. and Ugwu C.
Computer Science Department, University of Port-Harcourt, Rivers State, Nigeria.
nneka2k@yahoo.com, austinechidumo@yahoo.com
Abstract
The advances in electronic technology have created opportunities for new instructional designs to be
developed. Since Knowledge is expanding at a tremendous rate, these designs make it easy to impact the
necessary knowledge (both practical) on people to enable them survive in a competitive environment. The
Collaborative E-Learning Environment as an important group learning mode sets a goal of convenience and
an adaptation into our technologically advanced society. It emphasizes the communication and information
sharing among group members. Implementing this collaborative e-learning environment on a Linux thin-client
system makes it possible for this environment to be available in most schools and companies because the Linux
thin-clients are less expensive than other conventional computing systems. Developing a Collaborative ELearning Environment on a Linux Thin-Client System provides a means of delivering an improved quality of
education in our society. This paper involves the setting-up of Linux Thin-Client system, installing the
appropriate applications necessary for this environment, and developing the interactive portal that will enable
registered users to have access to the appropriate collaborative tools.
Keywords: Bandwidth, Collaborative learning, E-learning, Linux, Thin-client
1.0
Introduction
Students learn best when they are actively
involved in the process. Researchers report that,
regardless of the subject matter, students
working in small groups tend to learn more of
what is taught and retain it longer than when the
same content is presented in other instructional
formats. Students who work in collaborative
groups also appear more satisfied with their
classes.
Learning is enhanced when it is more like a
team effort than a solo race. Good learning, like
good work, is collaborative and social, not
competitive and isolated. Working with others
often increases involvement in learning. Sharing
one’s ideas and responding to other peoples
ides,
improves
thinking
and
deepen
understanding [3].
The term collaborative learning has been used
in two different senses. In one sense, some have
treated collaborative learning as a distinctive
form of socially based learning that is
fundamentally different from prevailing
.
psychological formulations. Another way to
think about collaborative learning is not as a
type of learning at all, but rather as a theory of
instruction [4].
[2] defined collaborative learning as "a reacculturative process that helps students become
members of knowledge communities whose
common property is different from the common
property of the knowledge communities they
already belong to ".Collaborative E-Learning
Environment” is a collaborative, convenient,
adaptive, and productive learning environment
which allows different participants to communicate
and coordinate with each other in a productive
manner. Productivity and efficiency is obtained
through synchronized communication between the
different coordinating partners [8]. Within this
process
of
synchronized
communication,
coordination can be accomplished by voice
communication through video/voice conferencing,
chat tools, share desktop and share applications [6].
Consequently, understanding the ideas and the
techniques behind synchronized communication
can be of great significance in the development of a
Collaborative E-Learning Environment.
In this paper, a collaborative e-learning
environment was developed on a Linux thin-client
system, to allow communication between students
and tutor(s). The operating system used is Fedora
Core Linux. 2xThinClientServer software was
installed on all the systems, and it was configured
as a master server. A username and password was
supplied to enable connection from the master
server to MySQL database. After which the
software that contains all the tools needed in the
collaborative environment was installed. This
software is the Global Communications Network
(GCN) software.
The Mozilla Firefox web browser was set on the
server to popup with the URL to the portal services.
This was achieved by writing a bash shell script
that ran at systems start-up. A user must log in to
the portal before he/she can have access to the
lectures going on, lecture notes, and interact with
other students that are equally logged on to the
collaborative e-learning environment.
2.0 Review Of Relevant Concept
The Linux thin-client technology is the system
upon which collaborative e-learning environment is
developed. Linux thin-clients have proven to be
extremely reliable because tampering with settings
are virtually non-existent, and in addition, an
educational institution will also gain more control
over how their students are using computing
resources and access to data. Linux thin-client is
distributed under the General Public License
(GPL), meaning it is free.
A thin-client (sometimes also called a lean or
slim client) is a client computer or client software
in client-server architecture networks which
depends primarily on the central server for
processing activities, and mainly focuses on
conveying input and output between the user and
the remote server. Thin-client can also be said to be
a generic name for client machines on networks that
deliver applications and data via a centralised
computing model with the majority of processing
carried out by the server(s). This is also known as
server-based computing [1]. Typically, the thinclient terminal sends key strokes and mouse clicks
to the server and the server sends screen updates to
the terminal device Because of its flexibility and
.
cost effectiveness, thin client has been empowering
schools, businesses, and organizations all over the
world. Thin-client technology has high data
security, low-maintenance, and terminals are less
likely to be affected by virus because by default
users are unable to tamper with settings [5].
Moreover, the entire system is simpler and easier to
install and to operate.
Linux Terminal Server Project (LTSP) is a
free and open source add-on package for Linux that
allow many people to simultaneously use the same
computer. Applications are run on the server with a
terminal known as a thin-client (also known as an X
terminal) handling input and output. Generally, they
are low-powered, absence of a hard disk; produce
less noise than standard PCs, less prone to problems
because they have fewer moving parts and no
installed applications, as such producing a pleasant
learning environment.
Server-Based Computing (SBC) is a technology whereby
applications are deployed, managed, supported and executed
on the server and not on the client. Only the screen information
is transmitted between the server and client. This architecture
solves the many fundamental problems that occur when
executing the applications on the client itself.
SBC environments hardware and software
upgrades, application deployment, technical
support, data storage and backup are simplified
because only the servers need to be managed. Data
and applications reside on a few servers rather than
on hundreds or thousands of clients [7].
3.0
Problem Definition
It is important to note that the economic survival
and development of a nation depends on the
productive capacity of the citizens, which depends
on the kind of training and education the
individuals receive from the educational sector in a
nation. The problem of poverty has been used in
most societies as an excuse and a reason why
quality training and education is not given to
students in most schools and organizations. This
paper provides a way of providing information,
bridge communication gap between people, and
also training at a cheaper and less expensive rate.
4.0 Design For The Collaborative E- Learning Environment
Figure 1: Design for Collaborative E-Learning Environment
Figure 1 illustrates a design for this Collaborative
E-Learning Environment.
The control computers (servers) consist of the
following: mail server, print server, Domain Name
.
System (DNS) server, Dynamic Host Configuration
Protocol (DHCP), Network-Attached Storage
(NAS), and the server that serves the thin-clients
with the user applications.
Client: These are the students, administrative
staffs, and academic staffs. Also the tutor can be an
e-tutor.
DHCP Sever: This is the Dynamic Host
Configuration Protocol sever. It is used basically to
dynamically allocate IP addresses to systems whose
IP addresses are not statically configured. It is used
mainly for fat clients that will be attached to the
network, that is, in cases where some students or
staff come with a laptop.
DNS Sever: The Domain Name Service/System
(DNS) sever is used to translate domain names into
IP addresses.
Print Server: This server manages the shared
printers.
Network Attached Storage (NAS): This is the file
server, and it is dedicated to file storage. This is
mainly where the electronic books and research
materials will be stored.
Domain and Cooperate Mail Server: This server
manages the domain and cooperate mail of staffs.
Firewall and Router: This server is the gateway
which controls asses to and from the network.
5.0 Developing The Collaborative
E-Learning Environment
After choosing the server hardware (the control
computer), the Fedora Core Linux OS distribution
was installed. After which the 2x ThinClientServer
software was installed.
The 2X ThinCientServer is configured before
starting up its services. This is done by opening the
terminal as root (super user) and running the
following commands:
2Xthinclientserver_conf.sh
In the first part, the configuration script detects
whether all necessary tools are present on the
system. Then it ‘backs up’ the current configuration
files for safe keeping. The license agreement is then
reviewed and accepted to continue.
Next, the network and security setup is probed. The
total number of network interfaces installed on the
machine is shown.
.
Figure 2: Master or Slave Menu
Since our server was configured as a master server,
there was a prompt for a username and password to
connect to the local MySQL database. The
following information was also asked for:
Master server host name or IP: This is used by
the slave servers, to communicate with the master
server. Please note that hostnames are resolvable by
all slaves. Internal IP address, as seen by the thin
clients: This IP is used by the thin clients to
communicate with the server.
After entering all the information, the details
were reviewed before proceeding with the
configuration. After this phase, individual services
were configured. The Apache and MySQL services,
including all other services are started and the script
exits. These lead to the configuration of the
consoles and settings for direct connection to the
terminal server. Settings were applied to individual
thin client, by username, by group (effectively
capturing all the members of the group).
The next step was installing the software that
was used in the collaborative environment. This
software is the Global Communications Network
(GCN) software. It contained all the tools that were
needed in the environment. These tools consist of:
E-mail, Chat Rooms, Private Messages, Games, a
Web Browser, Popup-blocker, Profiles, Message
Boards, Desktop Sharing, MSN, Yahoo, Voice Chat
Rooms,
Video
Conferencing,
Language
Translation, File Transferring, Whiteboards and a
Media Player.
The interactive portal was created using the
Hyper Text Markup Language (HTML). This portal
enabled the users of the collaborative e-learning
environment to have access to the needed tools.
1
6.0 Implementation, Methodology and
Updates
After 2X ThinClientOS has booted from the
client’s hard disk, it obtains the IP address of 2X
ThinClientServer from the network settings
returned by the 2X DHCP Helper Service.
Figure 3: 2X ThinClientOS booting up
It now prompts for the username and password.
Figure 4:Log on screen
Acknowledgement
We wish to acknowledge the contributions of
Miss Enemugwem J. Silverline of Department of
Computer Science, University of Port-Harcourt.
.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
Becta, 2007. http://www.becta.org.uk/publications (Accessed November 2010)
Brufee, K. 1993. Collaborative Learning. Baltimore, Johns Hopkins University Press
. UK. Pp24-32
Chickering, A. W., and Gamson, Z. F. 1987. Seven Principles for Good Practice in
Undergraduate Education, Wingspread Journal, vol, 9. No, 2. Pp 232-241.
Dillenbourg, P. 1999. Collaborative Learning, Cognitive and Computational
Approaches, Oxford, Pergamon UK. Pp 53-57
Richards, D. 2007. Linux Thin Client Networks, Design and Deployment, Packet
Publishing, Birmingham. Pp 33-41
Roschelle, J., and Teasley, S. 1995. The Construction of Shared Knowledge in
Collaborative Problem Solving in Computer Supported Collaborative Learning, Claire O'Malley, Berlin, SpringerVerlag USA. Pp 42-54.
Server based computing, 2005. Retrieved from
http://www.2x.com/whitepapers/WPserverbasedcomputing.pdf (Accessed April 2010)
Tessmer, M., and Harris, D. 1992. Analyzing the instructional setting, Kogan publishers,
London. Pp 67-73
.
An assessment of Internet Abuse in Nigeria
M.E Ezema*, H.C. Inyama+
Computer Science Department, University Nigeria Nsukka
Email: ezemamodesta@yahoo.com
+Department of Computer and Electronics Engineering, Nnamdi Azikiwe University Awka
Anambra State Nigeria, Email : drhcinyiama@gmail.com Phone: 08034701121
Abstract
As Internet use has proliferated worldwide, there has been debate whether some users develop disturbed
patterns of Internet use (i.e., Internet abuse). This article highlights relevant literature on Internet abuse in
Nigeria. Is the addiction paradigm appropriate for Internet use? Is behavior that has been labeled Internet
abuse symptomatic of other problems such as depression, sexual disorders, or loneliness in Nigeria? What are
alternative explanations for this phenomenon? Is there adequate research to support Internet abuse as a
distinct disorder?
Key words: Internet, Packet Switching, World Wide Web, Computer Crime, Cyber-bullying Malware
Introduction
The Internet was the result of some visionary
thinking by people in the early 1960s that saw great
potential value in allowing computers to share
information on research and development in
scientific and military fields. J.C.R. Licklider of
MIT first proposed a global network of computers
in 1962, and moved over to the Defense Advanced
Research Projects Agency (DARPA) in late 1962 to
head the work to develop it. Leonard Kleinrock of
MIT and later UCLA developed the theory of
packet switching, which was to form the basis of
Internet connections. Lawrence Roberts of MIT
connected a Massachusetts computer with a
California computer in 1965 over dial-up telephone
lines. It showed the feasibility of wide area
networking, but also showed that the telephone
line's circuit switching was inadequate. Kleinrock's
packet switching theory was confirmed. Roberts
moved over to DARPA in 1966 and developed his
plan for ARPANET. These visionaries and many
more left unnamed here are the real founders of the
Internet
What is Internet Abuse?
Defining Internet abuse is the first challenge,
and creating an organization wide acceptable use
policy (AUP) is the first step in the definition.
Internet abuse refers to improper use of the
internet and may include: computer crime, cyber
bullying, spam and malwares. An acceptable use
policy defines what constitutes Internet abuse in an
organization. Acceptable Internet behaviour in one
organization may be unacceptable in another, so the
.
acceptable use policy is a highly customized
policy, based on the organizational mission. The
organization determines what lines will be drawn
when it comes to Internet abuse. The amount of
resources and information the Internet contains is
astounding. With the help of information collected
on the net, people gain vast knowledge. Parents and
children together can work to make the Internet a
positive experience. However, some people can
misuse this wonderful knowledge bank and with no
rules or regulations, can discover surreptitiously
how to commit crimes, see things they ought not to
see and chat with people of questionable
character[1]. Thus parents ought to take precautions
to see that their children do not abuse the internet
access. On the other hand there are people who use
the internet for nefarious activities and they strike
to
win
converts
among
the
unwary.
INTERNET ABUSE
.
At home
In a friends
house
In the a cyber
cafe
In the cafeteria
At the market
place
Fig 1: Review of the internet abuse in Nigeria
Close monitoring and forbidden access
The location of a computer may make a huge
difference in the type of Web content one reads and
surfs. If possible, computers should be in the office,
living room, family room or some high traffic area
so that one can always monitor the internet access.
This will restrict abuse of the internet access at any
given moment since someone may approach the
computer while internet abuse is on going, hence
people will be more cautious and careful of their
online activities. On the other hand the cases of
night browsing, parents should not allow their
children to go to places they do not have adequate
trust on what their child may be doing or likely to
be doing there at night. The highest of all is
disciplined parents have to tell their children the
implications of certain actions like watching bad
films, discussing online with people you do not
know their family background very well so that
even if their parents are not at home with them they
will be limited with what they do with the internet.
Standard Internet Safety
Another important thing that one needs to know
is standard internet safety [3]. The key to a
.
successful
acceptable
use
policy (AUP)
implementation in most organizations is similar to
other policy development issues in the workplace.
There must be “buy-in” from the “top-down”, in
other words, the leaders of the organization must
agree to the principles of the AUP and endeavour to
push that policy down to the directors, managers
and supervisors within the organization. The most
critical stage of AUP development is dependent on
upper management “buy-in” and their willingness
to demonstrate the importance of this policy to the
rest of the organization.
It is very essential for one to know about
computers and be familiar with the World Wide
Web. Nothing can be more intimidating than a child
knowing more about computers and internet than
their parents, and often this is what happens with
today's parents who probably know very little about
internet compared to their children. Thus consider
this aspect no one can know if something is amiss
with a child while being totally repugnant if you do
not know or understand the child's online
activities
The Internet has become an invaluable resource
in the workplace, the world's biggest reference
library, social media centre, and pornography outlet
is now only a click away. This availability presents
a significant risk factor for employer liability and
costs employers thousands of hours in productivity
each day. Monitoring employee Internet use is one
way to reduce employer liability, and whether or
not you agree with the principles behind Internet
monitoring, many employers agree that it is a
necessary evil [2]. Internet abusers range from
upper management employees in private offices
viewing hardcore pornography, to the department
assistant in a cubicle that spends 3 hours a day
using Facebook, doing online shopping, making
travel arrangements, and paying bills through the
company Internet. Internet abuse is endemic in the
workplace and organizations are being forced to
face the problem head on, or suffer the
consequences.
Among the many consequences of Internet
abuse is a loss of productivity and scores of
litigation issues such as sexual harassment, hostile
work environment and discrimination. Monitoring
Employee Internet access is one way that an
organization can limit its liability.
Holding a series of Internet workshops with
employees of an organization is one way to
introduce new acceptable use policy. As an
educational session, an Internet workshop can
address the sensitive issues surrounding Internet
abuse in an open forum where employees can ask
questions and provide input in a nonconfrontational setting.
During the Internet workshop, the organization
can begin to educate the employees about Internet
abuse and give them a chance to re-evaluate their
Internet habits at work. It is important to be as open
as possible with employees regarding chosen
methodology for enforcing the AUP
For example, if the organization has decided to
employ Internet blocking technologies, the AUP
should define the specific types of websites that
will be blocked, for example, many organizations
block pornography, “gross depictions” and “hate”
websites. Discussing the types of websites the
organization has decided to block and answering
questions regarding the reasons for blocking will
reinforce the organizational mission, and
.
demonstrate the types of websites that are
inappropriate within an organization.
If an organization is going to monitor and report
on employee Internet access, the workshop will
give one a chance to show the employees what the
Internet reports look like, and discuss the
circumstances in which they will be used. Taking
the mystery out of what the organization is planning
in regards to Internet monitoring and blocking will
reduce employee speculation and set new
expectations throughout the organization
Problems with Internet Monitoring
The technical aspects of blocking website access
and monitoring employee Internet access are not
without problems. The software for blocking
websites has advanced tremendously over the past 5
years; however, there are still problems with
blocking “all” inappropriate websites and blocking
websites that you did not intend to block. No
system is perfect and one will need assistance from
a selected software or hardware vendor in addition
to information systems department. If possible, it is
always better to meet, in person, with the vendor
representatives prior to the purchase of any Internet
monitoring software. Voice your concerns with the
vendor and secure “after sale” support with the
vendor help desk. If you have an information
systems department, one should make sure they are
involved from the start of the project to help
address any technical problems that the new system
could bring.
Monitoring Employee Internet Access - The
People Side
Outside of the technical issues that will occur,
the people side of Internet monitoring can be the
most problematic of all. Even with the
dissemination of information given at the Internet
workshop and taking great care during policy
development, some employees will, inevitably feel
that Internet monitoring is unfair. Given this fact, it
is of the utmost importance that the Internet reports
are accurate, beyond question. Even if they are
correct, there are still issues to consider. The
scenarios listed below are examples of how
employees could react if they are confronted with
the accusation of Internet abuse.
Moreover, the excuses below may be completely
accurate and good explanation by the accused.
"It wasn't me!"
It is always possible that some other person was
on the accused employee’s computer surfing the
Internet. Once a user steps away from the computer,
anything can happen. Another person sits down and
starts using the computer logged in as the accused,
everything they do on the Internet is recorded under
somebody else's name. One suggestion is to have
the user lock their computer before leaving for an
extended period of time; this will reduce the
chances of misidentification of the Internet abuser.
"They have my password"
This is a similar situation to the one mentioned
above. If I have a user's password, I could log-in as
the user and all of my Internet access would be
attributed to them. How they got the password is
another issue entirely, however the user makes a
good point and has a potentially valid excuse for an
Internet report that shows abuse.
.
helpful in reducing employer liability and
"The Internet Report is Wrong"
improving employee productivity. Developing an
This can occur if the monitoring software is
acceptable use policy to outline acceptable Internet
setup incorrectly or if there are network issues
behaviour in an organization is the first step in the
causing identification problems. This is another
process. To implement this policy successfully, the
reason why one will want information systems
policy must be supported by upper, mid, and line
department involved from the start and technical
level managers. The organization should
support from the vendor who sold the Internet
endeavour, with enthusiasm, to educate the
monitoring solution. Defending an Internet report
employees of the organization about Internet abuse
that shows abuse is difficult when you do not
and share the organizations plans to monitoring use
understand how the technical aspects of Internet
and block inappropriate websites.
monitoring work.
Prior to purchasing a software or hardware
Internet reporting is not an exact science, the
solution for Internet monitoring and blocking, a
reports could be wrong, and the person accused of
vendor should be selected and invited into the
Internet abuse may be completely innocent. The
organization to explain the technical problems that
key is to research the potential offender and look
can occur with Internet monitoring and blocking
into their history. People who abuse the Internet
technologies. During this vendor selection process,
usually have a history of doing so, so look into their
it is very important to include information systems
past Internet use first and then look at the Internet
department and other technical staff. Arranging
records on their computer. In short, do a “reality
after-sale support with the vendor of choice is
check”. Too often we take technology for its word
highly recommended.
and fail to look on the human side for insight that
Finally, there is the people side of the problem.
may confirm or make us question our suspicions.
Internet monitoring and blocking are only as good
This practice will help reduce the number of errors
as the software and hardware solutions that are
that could be made during the investigation of
developed. There are many ways that these
Internet abuse, and help the employer maintain their
solutions can fail, so doing a thorough investigation
credibility.
prior to accusing an employee of Internet abuse is
Internet abuse is a fact of life in most large
also highly recommended.
organizations today. Monitoring employee Internet
use and employing blocking technologies can be
___________________________________________________________________________________
References
[1]
[2]
[3]
Acier, Didier and Laurence Kern. “Problematic Internet use: Perceptions of Addiction
Counselors.” Computers and Education. May 2011, Vol. 56: 983-989.
Block, Jerald. “Issues for DSM-V: Internet Addiction.” American
Journal of Psychiatry. 2008, Vol. 165 No. 3: 306-307.
Internet Abuse, www.buzzle.com/editorials/1-13-2005-64163.
.
Instructions for Authors
WAJIAR
provides a multidisciplinary forum for the publication of original research and technical papers, short
communications, state-of-the-art developments in Information and communications technology, bio-medical
informatics, computers, mathematics, control and information technology, physics, industrial electronics and
engineering, Industrial chemistry, general technology, rocketry, space and astronomy, earth science, oceanography and
the environment with special emphasis on the application of communications and information technology to these fields
of importance in our modern civilization. The journal will review papers on advances, techniques, practice and
application of information and communications technology in these areas.
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
Economics, Statistics and Mathematic
Information Technology & Applications
Electronics and Computer Engineering
Computer networks, Satellite & communications Engineering Research
Industrial Computer applications Today
Computer and Electro-Mechanical Maintenance
GPRS and Remote Sensing
Robotics ,Telemedicine & Remote Medical Monitoring
Artificial Intelligence & Expert Systems Development & Applications
Developments in Forensic Science Research
Information Assurance & Network Security: The African Dilemma
Space Information Systems & Rockery : which way for Africa
Video and Teleconferencing Innovations for Deployment in Africa
Telecommunications Technology & 4G Applications
Biotechnology
Agriculture , Food Technology & Processing Environmental Technology & Impact Analysis
E-Waste Management & Environmental Protection
Management Science & Operations Research
Wireless Technology, GSM and 4G Applications
Alternative ,Grid and Green Energy Solutions for Africa
Converting Academic Research in Tertiary Institutions for Industrial Production
WAJIAR is an international Journal of Science and Technology and is published quarterly in February, May, August and
December.
.
Key Objectives:
provide avenue for the advancement, dissemination of scientific and technology research in the fields of science and
technology with special emphasis on the deployment of Information & Communications Technology as an enabler
and solution tool.
Versions:
Two versions of the journal will be published quarterly (electronic version posted in the web and hardcopy version.
Submission
2
Typescript and soft copy in MS-word format should be submitted by email to: Info@wajiaredu.com. Typescripts
should be written in English, double-spaced and single-sided on either Quarto or A4-sizeed sheets, and should be
numbered beginning with the title page. The first page of an article should contain:
i)
The title of the paper, the name(s) and address(s) of the author(s)
ii)
A short title not exceeding 45 letters with space
An abstract of 50-200 words should follow the title page on a separate sheet.
2.1 Text
Headings and subheadings for different sections of the paper (e.g. Introduction, Methods, Results, and Discussions)
should be clearly indicated. Units of measurement, abbreviations and symbols should follow the international system of
units (SI), Equations and formulae should be typewritten or word-processed. Equations should be numbered
consecutively with Arabic numerals in parentheses on the right-hand side of the page. Special symbols should be
identified in the margin.
Tables and Figures
Tables should be given short informative titles and should be numbered consecutively in Arabic numerals. Tables will be
reproduced directly from the typed or word-processed softcopy. Figures in a form suitable for reproduction should be
submitted separately from the text as original drawings in Indian ink, or as high contrast sharp photographs on glossy
paper. Lettering on figures should be proportional to the size of the figure to ensure legibility after reduction. On the back
of each figure, the orientation of the figure and the author and figure number should be cited consecutively in the text and
a list of short descriptive captions should be provided at the end of the paper.
2.2 References
References should be cited in the text using a number in square brackets in order of appearance. The references are listed
at the end of the paper in numerical order.
General Information
Proofs: Proofs will be sent to the nominated author to check typesetting accuracy. No changes to the original manuscript
are accepted at this stage. Proofs should be returned within seven days of receipt.
.
Reprints: reprints may be ordered (pre-paid) at prices shown on the reprint order form which will be sent to the author
together with the proofs.
Page Charge: Please note that there will be a page charge of N20,000 or $100 (Twenty thousand naira or One Hundred
American Dollars only) for each article accepted for publication in the journal.
Annual Subscription is pegged at N20,000 or $125 per annum for four volumes. Electronic copies will be free access at
our website. Subscription for the printed version will be completed via our Website www.wajiaredu.com. Subscriptions
can also be sent via the journal’s email address: www.wajiaredu.com/webmail / info@wajiaredu.com .ISSN: 2276-
9129
.
& Academic Research
Publications Office:
9-14 mbonu ojike Street
Ikenegbu, Owerri, Nigeria
International office::
255 North D Stree
San Bernardino, CA 92401
909.884.9000
Tel: 234 81219 41139
www.wajiaredu.com
Editor-in-Chief:
Prof. Oliver E. Osuagwu, PhD, FNCS, FBCS CITP, MIEEE, MACM
Editorial Board: Prof Tony B.E. Ogiemien, PhD, BL, (USA), Engr. Prof E. Anyanwu, Ph.D, FNSE, Prof. G. Nworuh, PhD, Dr. B. C. Ashiegbu, PhD, Prof. C.O.E. Onwuliri, PhD, FAS ,
Prof .E. Emenyionu, PhD, (Connecticut USA,) Prof. E.P. Akpan, Ph.D, Engr. Prof. C.D. Okereke, Ph.D, Prof. B.E.B. Nwoko, Ph.D, Prof. N.N. Onu, PhD, Prof M.O. Iwuala, PhD, Prof C.E.Akujo, PhD,
Prof. G. Okoroafor, PhD, Prof Leah Ojinna, Ph.D (USA), Prof. O. Ibidapo-Obe, PhD, FAS., Prof. E. Adagunodo, PhD, Prof. J.C .Ododo, PhD, Dan C. Amadi, PhD(English), Prof.(Mrs) S.C. Chiemeke, PhD,
Prof (Mrs) G. Chukwudebe,PhD, FNSE, Dr. E.N.C. Okafor, PhD, Dr (Mrs) I. Achumba, Dr. T. Obiringa, PhD, Dr. S. Inyama, PhD, Prof. C. Akiyoku, PhD, Prof. John Ododo, PhD, Prof. E. Nwachukwu,
Ph.D, FNCS, Dr. S. Anigbogu, PhD,FNCS, Prof. H. Inyama, PhD, FNSE .Prof. B.N.. Onwuagba, PhD, Prof J.N. Ogbulie, PhD
_________________________________________________________________________________________________________________________
Subscription Application Form
Name of Subscriber:_____________________________________________________________
Institutional affiliation:___________________________________________________________
Mailing Address to send copies of journals __________________________________________
_____________________________________________________________________________
Email Address:_____________________________Tel:_________________________________
Version of Journal requested: (a) hard copy N20,000 or $125
(b) electronic N10000 or $63*
Method of payment:
Wire Transfer
Visa Card
Master Card
Direct payment into our Diamond Bank
Account No:. 0024189096 Douglas Rd branch, Owerri
Bank Draft
Tick
All subscription Application should be forwarded to: info@wajiaredu.com
Subscribers Signature: _______________________ Date: DD/MM/YYYY___/____/___
*WWW.WAJIAREdu.com is free access. Electronic copies in form of e-book requested is what is mean here by “electronic”.

West African Journal of Industrial & Academic Research ISSN: 2276-9129

Transcription

Similar documents

African customers, African partners PM Zone: Foreign Customers

pharmacy - Cooper

African American Heritage Festival

Franz Schubert Impromptus, Moments Musicaux, Klavierstücke for

Akpe Brochure Short

Emerging African News

KICKS AND GEAR RUNS THIS KICKS AND GEAR

Africa Political Map

Markham public library

mabula private game reserve - 2016

West African Journal of Industrial &amp; Academic Research ISSN: 2276-9129

Transcription

Similar documents

West African Journal of Industrial & Academic Research ISSN: 2276-9129