资源描述:
《Modern Robust Statistical Methods》由会员上传分享,免费在线阅读,更多相关内容在学术论文-天天文库。
1、ModernRobustStatisticalMethodsAnEasyWaytoMaximizetheAccuracyandPowerofYourResearchDavidM.Erceg-HurnUniversityofWesternAustraliaVikkiM.MirosevichGovernmentofWesternAustraliaClassicparametricstatisticalsignificancetests,suchasnormality)areviolated.Themiscomputat
2、ionofpvalues,analysisofvarianceandleastsquaresregression,arecoupledwiththeinaccurateestimationofeffectsizesandwidelyusedbyresearchersinmanydisciplines,includingconfidenceintervals,canleadtosubstantiveerrorsinthepsychology.Forclassicparametricteststoproduceaccu
3、-interpretationofdata.Severalprominentstatisticiansandrateresults,theassumptionsunderlyingthem(e.g.,nor-researchershavedescribedtheuseofclassicparametricmalityandhomoscedasticity)mustbesatisfied.Theseas-statisticsinthefaceofassumptionviolationsasinvalidsumptio
4、nsarerarelymetwhenanalyzingrealdata.The(e.g.,Kezelmanetal.,1998;Leech&Onwuegbuzie,2002;useofclassicparametricmethodswithviolatedassump-Wilcox,2001;Zimmerman,1998).tionscanresultintheinaccuratecomputationofpvalues,Modernrobuststatisticalproceduresexistthatcane
5、ffectsizes,andconfidenceintervals.Thismayleadtosolvetheproblemsinherentinusingclassicparametricsubstantiveerrorsintheinterpretationofdata.Manymod-methodswhenassumptionsareviolated.Manymodernernrobuststatisticalmethodsalleviatetheproblemsinher-statisticalproced
6、uresareeasilyconductedwithwidelyusedentinusingparametricmethodswithviolatedassumptions,softwaresuchasSPSS,SAS,andR.Despitetheadvan-yetmodernmethodsarerarelyusedbyresearchers.Thetagesofmodernmethodsandtheeasewithwhichtheseauthorsexaminewhythisisthecase,arguing
7、thatmostprocedurescanbeconducted,theyarerarelyusedbyresearchersareunawareoftheseriouslimitationsofclas-researchers.Inthefirstpartofthisarticle,weexaminewhysicmethodsandareunfamiliarwithmodernalternatives.Athisisthecase.Wearguethatmostresearchersareunawarerange
8、ofmodernrobustandrank-basedsignificancetestsofthelimitationsofclassicmethodsanddonotrealizethatsuitableforanalyzingawiderangeofdesignsisintro-modernalternativesexist.Inthesecondhalfofthear