SAMPLE MEANS 8/4/2014
Transcription
SAMPLE MEANS 8/4/2014
8/4/2014 SAMPLE MEANS Section 7-3 Estimating a Population Mean 1. Formanypopulations,thedistributionof samplemeans ̅ tendstobemore consistent(withlessvariation)thanthe distributionsofothersamplestatistics. 2. Forallpopulations,thesamplemean ̅ is anunbiasedestimatorofthepopulation meanµ,meaningthatthedistributionof samplemeanstendstocenteraboutthe valueofthepopulationmeanµ. POINT ESTIMATE COMMENT PointEstimate: Thesamplemean ̅ isthe bestpointestimate (orsinglevalueestimate) ofthepopulationmean . Itisrarethatwewanttoestimatethe unknownvalueofapopulationmeanbutwe somehowknowthevalueofthepopulation standarddeviation .Therealisticsituationis that isnotknown.(Webeginthissectionby consideringthismorerealisticscenario.) When isnotknown,weconstructthe confidenceintervalbyusingtheStudent distributioninsteadofthestandardnormal distribution. ASSUMPTIONS FOR CONFIDENCE INTERVAL OF MEAN WITH σ NOT KNOWN THE STUDENT t DISTRIBUTION 1. Thesampleisasimplerandomsample. 2. Eitherorbothofthefollowingconditions aresatisfied: • Thepopulationisnormally distributed • n >30 Ifapopulationhasanormaldistribution,then thedistributionof ̅ isaStudentt distribution forallsamplesofsize .TheStudent distributionisoftenreferredto asthe distribution. 1 8/4/2014 FINDING THE CRITICAL VALUE DEGREES OF FREEDOM Findingacriticalvalue / requiresavaluefor thedegreesoffreedom (ordf).Ingeneral, thenumberofdegreesoffreedomfora collectionofsampledataisthenumberof samplevaluesthatvaryaftercertainrestraints havebeenimposedonthedatavalues.Forthe methodsofthissection,thenumberofdegrees offreedomisthesamplesizeminus1;thatis, degreesoffreedom 1 MARGIN OF ERROR ESTIMATE OF µ (WITH σ NOT KNOWN) / CONFIDENCE INTERVAL ESTIMATE OF THE POPULATION MEAN μ (WITH σ NOT KNOWN) ∙ ̅ where(1 )istheconfidenceleveland has 1 degreesoffreedom. 1. 2. Verifythatthetworequiredassumptionsaremet. With unknown(asisusuallythecase),use 1 degreesoffreedomandrefertoTableA‐3tofindthe criticalvalue / thatcorrespondstothedesired confidenceinterval.(Fortheconfidencelevel,refer to“AreainTwoTails.”) 3. Evaluatethemarginoferror 4. Findthevaluesof ̅ and ̅ .Substitutethese inthegeneralformatoftheconfidenceinterval: ̅ ̅ . Roundtheresultusingthesameround‐offruleon thefollowingslide. / ̅ / CONSTRUCTING A CONFIDENCE INTERVAL FOR μ (σ NOT KNOWN) 5. Acriticalvalue / canbefoundusingTableA‐3which isfoundonpage586,insidethebackcover,andonthe FormulasandTablescard.Ifthetabledoesnotinclude thenumberofdegreesoffreedomthatyouneed,you could • usetheclosestvalue • beconservativeandusingthenextlowernumber ofdegreesoffreedom • interpolate.Forexample,ifyouhave55degrees offreedom,youcouldfindthemeanofthe criticalvaluesfor50and60. Tokeepthingssimple,wewillusetheclosestvalue. ∙ where / ∙ ROUND-OFF RULE FOR CONFIDENCE INTERVALS USED TO ESTIMATE μ 1. Whenusingtheoriginalsetofdata to constructtheconfidenceinterval,round theconfidenceintervallimitstoonemore decimalplace thanisusedfortheoriginal dataset. 2. Whentheoriginalsetofdataisunknown andonlythesummarystatistics , ̅ , areused,roundtheconfidenceinterval limitstothesamenumberofplaces as usedforthesamplemean. 2 8/4/2014 FINDING A CONFIDENCE INTERVAL FOR µ WITH TI-83/84 1. 2. 3. 4. 5. 6. 7. 8. 9. Select STAT. Arrow right to TESTS. Select 8:TInterval…. Select input (Inpt) type: Data or Stats. (Most of the time we will use Stats.) Enter the sample mean, x. Enter the sample standard deviation, Sx. Enter the size of the sample, n. Enter the confidence level (C-Level). Arrow down to Calculate and press ENTER. PROPERTIES OF THE STUDENT t DISTRIBUTION (CONTINUED) 2. 3. 4. 5. TheStudent distributionhasthesamegeneral symmetricbellshapeasthenormaldistribution butitreflectsthegreatervariability(withwider distributions)thatisexpectedwithsmallsamples. TheStudentt distributionhasameanof 0 (justasthestandardnormaldistributionhasa meanof 0). ThestandarddeviationoftheStudentt distributionvarieswiththesamplesizeandis greaterthan1(unlikethestandardnormal distribution,whichhasa 1). Asthesamplesizen getslarger,theStudent distributiongetsclosertothenormaldistribution. FINDING A CONFIDENCE INTERVAL FOR µ WITH TI-83/84 1. 2. 3. 4. 5. 6. 7. 8. 9. SelectSTAT. ArrowrighttoTESTS. Select7:ZInterval…. Selectinput(Inpt)type:Data orStats.(Mostof thetimewewilluseStats.) Enterthestandarddeviation,σ. Enterthesamplemean, . Enterthesizeofthesample,n. Entertheconfidencelevel(C‐Level). ArrowdowntoCalculate andpressENTER. PROPERTIES OF THE STUDENT t DISTRIBUTION 1. TheStudent distributionisdifferentfor differentsamplesizes(seeFigurebelow forthecases 3 and 12). ESTIMATING A MEAN WHEN σ IS KNOWN Requirements: 1. Thesampleisasimplerandomsample. 2. Eitherorbothoftheseconditionsaresatisfied:The populationisnormallydistributedor 30. ConfidenceInterval: ̅ ̅ wherethemarginoferror isfoundfromthefollowing: / ⋅ Note:Thecriticalvalue / isfoundfromTableA‐2(the standardnormaldistribution). SAMPLE SIZE FOR ESTIMATING µ ⁄ ∙ where zα/2 = criticalz scorebasedondesired confidencelevel E = desiredmarginoferror σ = populationstandarddeviation 3 8/4/2014 ROUND-OFF RULE FOR SAMPLE SIZE n Whenfindingthesamplesize ,iftheuseof theformulaonthepreviousslidedoesnot resultinawholenumber,alwaysincrease the valueof tothenextlarger wholenumber. CHOOSING THE APPROPRIATE DISTRIBUTION FINDING THE SAMPLE SIZE WHEN σ IS UNKNOWN 1. Usetherangeruleofthumb(seeSection3‐3) toestimatethestandarddeviationasfollows: range/4. 2. Startthesamplecollectionprocesswithout knowing and,usingthefirstseveralvalues, calculatethesamplestandarddeviations and useitinplaceofσ.Theestimatedvalueof canthenbeimprovedasmoresampledataare obtained,andtherequiredsamplesizecanbe adjustedasyoucollectmoresampledata. 3. Estimatethevalueofσ byusingtheresultsof someotherearlierstudy. CHOOSING BETWEEN z AND t Conditions Method σ notknownandnormallydistributed population or σ notknownand 30 UseStudent distribution σ knownandnormallydistributed population or σ knownand 30 Populationisnotnormallydistributed and 30. Usenormal( ) distribution Useanonparametric methodor bootstrapping FINDING A POINT ESTIMATE AND E FROM A CONFIDENCE INTERVAL Pointestimateofµ: ̅ upperconfidencelimit lowerconfidencelimit 2 Marginoferror: upperconfidencelimit lowerconfidencelimit 2 4