ResearchMethods,Instruments,&Computers1990,22
(5),486 -BOOKREVIEW Howmanysubjects?
Statisticalpoweranalysisinresearch STEPHENDUBINDrexelUniversity,Philadelphia,Pennsylvania Howmanysubjects?
StatisticalpoweranalysisinresearchByHelenaChmuraKraemer&SueThiemann.1987,NewburyPark,CA:SagePublications,Inc.120pp.ISBN:0-8039-2944-
8,$17.95. Itismostsatisfyingtofindastatisticalbookthatisinformativeandmathematicallyrigorous,yetverypleasantreadinginamanageableformat.HowManySubjects?
,byHelenaChmuraKraemerandSueThiemann,isjustsuchalittlegem.Thisstylishlyslimvolume(120pages,ofwhich20arechapterdividersorblanks)presentstheauthors'methodofsamplesizeestimationforawidevarietyofstatisticalprocedures,alongwithmuchusefulinstructionandmanyinsightsintogeneralexperimentaldesign. Researchinvestigatorsareincreasinglybeingcalledupontojustifythenumberofobservationsrequiredfortheirwork(cf.theNIHGuidefortheCareandUseofLaboratoryAnimals,AppendixD).Inadditiontoobviatingunnecessarydistresstosubjects,carefulplanningofexperimentspromoteseconomyandplaingoodscience.Ofthemanysourcesofinformationonstatisticallyappropriatesamplesize,manyaretask-specific,beinglimitedtoasingleorsmallnumberofstatisticalprocedures.Someareexhaustiveandexhaustingbecauseofthemassanddiversityofmaterial.Aregrettablefew-unfortunatelyattractivetothebeginnerbecauseoftheirsimplicity-handleonlyTypeI(0:)errorandignoretheveryimportantpowerside(TypeIIorI-(3)ofthesamplesizeissue. KraemerandThiemann'sbookbeginswithabriefbutlucidoverviewoftheconceptsofstatisticalhypothesistesting,criticaleffectsize,andpoweranalysisintheoverallcontextofresearchstudyplanning.Averyelegantfeatureofthebookistheauthors'innovationofasinglemethodofpowercalculationthatisappropriateforawidevarietyofdifferenttests.Notonlyisthisaesthetically StephenDubin,VMD,PhD,a"recoveringveterinarian,"isclinicalprofessorofbiomedicalengineeringandscienceatDrexelUniversityinPhiladelphia.Hisaddressis:DrexelUniversity,32nd&ChestnutSts.,Philadelphia,PA19104. appealing,butitalsomitigatestheeffortofunravelingthemathematicalbackground.Theauthorsdevelopthebasisoftheirmethodwithconcisemathematicalrigor.Inotherwords,theseriousstudentshouldbepreparedtoreadthereferences.Theythenapplythemethodtoawidevarietyofstatisticalprocedures,includingttests,analysisofvariance,variousformsofcorrelationandregressionanalysis,andnonparametric(binomial)testprocedures.Applicationofthemethodtocontingencytable(chi-square)testingispossiblebutabitplicatedthanitisfortheotherstatisticaltests.Thefinalpartofthebookcontainsasummarytableoftheformulasandthemastertables. Typically,onecalculatesavalueof0,whichisinturnaparameterforthecalculationof.:l.Thereaderthenusesthemastertablesordingtotherequiredlevelsofalphaandpowerandthevalueof.:lfortheirproposedstudy. Ratherthanprovideawidevarietyofexamples,KraemerandThiemannpresentasingle"threaded"examplepertainingtocoffeedrinkingandtheCornellMedicalIndex(eMI).Theyproceedtoillustratetheestimationofappropriatesamplesizeforeachofthestatisticalprocedureswithreferencetovariouswaysinwhichthesamecoffee/CMIstudymightbestructuredandanalyzed.Thisapproachcertainlydisqualifiesthebookasa"cookbook"inthemannerofSchaum'sOutlines.Itdoes,however,teachmuchmoreaboutexperimentaldesignthanjustsamplesize. Findingasamplesizeusingthisbookrequiresseveralstepsputation-sometimesinvolvingtranscendentalfunctions-andtheninterpolationwithinthemastertable.Thisestediousandpossiblyvulnerabletoerrorwhenthereaderwishestoaskmultiple''whatif"questionsaboutanexperimentaldesign.HowManySubjects?
isveryconcise;thenonexpertreaderwillhavetoreaditthroughwithcarefulattention.Everythingisthere,butonlyinoneplace!
Thisbookisanexcellentchoicefortheresearchinvestigatorwhowantsapleasantbutrigorousintroductiontopoweranalysis.Itisperhapsthesolesourcefortakingstudentsbeyondthecookbookstageofsamplesizedetermination. Copyright1990PsychonomicSociety,Inc. 486