11 Transformation Structural Equation Models Wit

See discussions, stats, and author profiles for this publication at: Structural Equation Models WithHighly Nonnormal and Incomplete DataArticle in Structural Equation Modeling A Multidisciplinary Journal · May 2014DOI: 10.1080/10705511.2014.937320CITATIONS0READS284 authors, including:Pengfei LiuJiangsu Normal University Xuzhou China3 PUBLICATIONS 0 CITATIONS SEE PROFILEZhaohua LuSt. Jude Children's Research Hospital18 PUBLICATIONS 79 CITATIONS SEE PROFILEAll in-text references underlined in blue are linked to publications on ResearchGate,letting you access and read them immediately.Available from: Pengfei LiuRetrieved on: 06 September 2016Thisarticlewasdownloadedby:[ChineseUniversityofHongKong]On:27July2015,At:01:02Publisher:RoutledgeInformaLtdRegisteredinEnglandandWalesRegisteredNumber:1072954Registeredoffice:5HowickPlace,London,SW1P1WGClickforupdatesStructuralEquationModeling:AMultidisciplinaryJournalPublicationdetails,includinginstructionsforauthorsandsubscriptioninformation::12Dec2014.Tocitethisarticle:PengfeiLiu,JiChen,ZhaohuaLu&XinyuanSong(2015)TransformationStructuralEquationModelsWithHighlyNonnormalandIncompleteData,StructuralEquationModeling:AMultidisciplinaryJournal,22:3,401-415,DOI:10.1080/10705511.2014.937320Tolinktothisarticle:(the“Content”)containedinthepublicationsonourplatform.However,Taylor&Francis,ouragents,andourlicensorsmakenorepresentationsorwarrantieswhatsoeverastotheaccuracy,completeness,orsuitabilityforanypurposeoftheContent.Anyopinionsandviewsexpressedinthispublicationaretheopinionsandviewsoftheauthors,andarenottheviewsoforendorsedbyTaylor&Francis.TheaccuracyoftheContentshouldnotberelieduponandshouldbeindependentlyverifiedwithprimarysourcesofinformation.TaylorandFrancisshallnotbeliableforanylosses,actions,claims,proceedings,demands,costs,expenses,damages,andotherliabilitieswhatsoeverorhowsoevercausedarisingdirectlyorindirectlyinconnectionwith,inrelationtoorarisingoutoftheuseoftheContent.Thisarticlemaybeusedforresearch,teaching,andprivatestudypurposes.Anysubstantialorsystematicreproduction,redistribution,reselling,loan,sub-licensing,systematicsupply,ordistributioninanyformtoanyoneisexpresslyforbidden.Terms&Conditionsofaccessandusecanbefoundat:AMultidisciplinaryJournal,22:401–415,2015Copyright©Taylor&FrancisGroup,LLCISSN:1070-5511print/1532-8007onlineDOI:10.1080/10705511.2014.937320TransformationStructuralEquationModelsWithHighlyNonnormalandIncompleteDataPengfeiLiu,1JiChen,2ZhaohuaLu,3andXinyuanSong21JiangsuNormalUniversity,Xuzhou,China,andChineseUniversityofHongKong2ChineseUniversityofHongKong3UniversityofNorthCarolina,ChapelHillAsusefulmultivariatetechniques,structuralequationmodelshaveattractedsignificantatten-tionfromvariousfields.Mostexistingstatisticalmethodsandsoftwareforanalyzingstructuralequationmodelshavebeendevelopedbasedontheassumptionthattheresponsevariablesarenormallydistributed.Severalrecentlydevelopedmethodscanpartiallyaddressviolationsofthisassumption,butstillencounterdifficultiesinanalyzinghighlynonnormaldata.Moreover,thepresenceofmissingdataisapracticalissueinsubstantiveresearch.Simplyignoringmiss-ingdataorimproperlytreatingnonignorablemissingnessasignorablecouldseriouslydistortstatisticalinfluenceresults.ThemainobjectiveofthisarticleistodevelopaBayesianapproachforanalyzingtransformationstructuralequationmodelswithhighlynonnormalandmissingdata.Differenttypesofmissingnessarediscussedandselectedviathedevianceinforma-tioncriterion.Theempiricalperformanceofourmethodisexaminedviasimulationstudies.Applicationtoastudyconcerningpeople’sjobsatisfaction,homelife,andworkattitudeispresented.Keywords:Bayesiansplines,MCMCalgorithm,missingdata,nonparametrictransformations,structuralequationmodelsInbehavioral,educational,andsocial-psychologicalresearch,themaininterestoftenfocusesnotonlyonobservedvariables,butalsoonlatentvariablesthatrequireassessmentbytwoormoreobservedvariables.Structuralequationmodelsarethemostimportanttoolsforexaminingtheinterrelationshipsamonglatentvariables.Overthepastseveraldecades,structuralequationmodeling(SEM)(Bollen,1989;Jöreskog&Sörbom,1996;Lee,2007)hasbeenextensivelyappliedtosummarizemultipleindicatorsofpsychological,social,ormedicalconstructs,aswellastoinvestigatehowoneormorelatentconstructsinfluenceothers.CommonlyusedsoftwaresuchasLISREL,EQS6,Mplus,andAMOSbroughtsignificantattentiontoSEMfromvariousfields.Themajorityofexistingstructuralequa-tionmodelsandrelatedsoftwarearebasedo