AMOS-SEM-analysis

AMOS–AnalysisofMomentStructuresHIVPreventionCenterUniversityofKentuckyRickZimmerman,OlgaDekhtyarOverviewOverviewofStructuralEquationModels(SEM)IntroductiontoAMOSUserInterfaceAMOSGraphicsExamplesofusingAMOSPredictorsofCondomUseusinglatentvariablesStructuralEquationModelsStructuralEquationModeling(SEM)AnextensionofRegressionandgeneralLinearModelsAlsocanfitmorecomplexmodels,likeconfirmatoryfactoranalysisandlongitudinaldata.StructuralEquationModelingAbilitytofitnon-standardmodels,databaseswithautocorrelatederrorstructurestimeseriesanalysisLatentCurveModels,databaseswithnon-normallydistributedvariablesdatabaseswithincompletedata.FamilyTreeofSEMT-testANOVAMulti-wayANOVARepeatedMeasureDesignsGrowthCurveAnalysisBivariateCorrelationMultipleRegressionPathAnalysisStructuralEquationModelingFactorAnalysisExploratoryFactorAnalysisConfirmatoryFactorAnalysisLatentGrowthCurveAnalysisNextWorkshop:November9Seeyouthere!StructuralEquationModeling(SEM)Exogenousvariables=independentEndogenousvariables=dependentObservedvariables=measuredLatentvariables=unobservedStructuralEquationGraphsObservedVariableLatentVariable.15Error:Loading.10:R2Example:CondomUseModelSEX1FRBEHB1ISSUEB1SXPYRC1ImpulsiveIDMC1RObservedVariablesLatentVariables.15LoadingsLegendIDMA1RIDME1RIDMJ1RRespondentSexCondomUseCondomattitudePeernormsaboutcondomsImpulsiveDecisionMakingObservedvariablesforImpulsivedecisionmakingExample:CondomUseModelSEX1FRBEHB1ISSUEB1SXPYRC1ImpulsiveIDMC1RObservedVariablesLatentVariables.15LoadingsLegendIDMA1RIDME1RIDMJ1RIndependentDependentDependentDependentIndependentDependentExample:CondomUseModelSEX1FRBEHB1ISSUEB1SXPYRC1ImpulsiveeSXYRC1efr1eissIDMC1Reidm4eidm2eidm2eidm1ObservedVariablesLatentVariables.15LoadingsLegendIDMA1RIDME1RIDMJ1RExample:CondomUseModelSEX1FRBEHB1ISSUEB1SXPYRC1ImpulsiveeSXYRC1efr1eissIDMC1Reidm4eidm2eidm2eidm1ObservedVariablesLatentVariables.15LoadingsLegendIDMA1RIDME1RIDMJ1RExample:CondomUseModelSEX1FRBEHB1ISSUEB1SXPYRC1ImpulsiveeSXYRC1efr1eissIDMC1Reidm4eidm2eidm2eidm1ObservedVariablesLatentVariables.15LoadingsLegendIDMA1RIDME1RIDMJ1R.11.38-.06-.10-.19-.15.13.53.49.69.67.15.05.28.24.48.45.03SEMAssumptionsAReasonableSampleSizeagoodruleofthumbis15casesperpredictorinastandardordinaryleastsquaresmultipleregressionanalysis.[“AppliedMultivariateStatisticsfortheSocialSciences”,byJamesStevens]researchersmaygoaslowasfivecasesperparameterestimateinSEManalyses,butonlyifthedataareperfectlywell-behaved[BentlerandChou(1987)]Usually5casesperparameterisequivalentto15measuredvariables.SEMAssumptions(cont’d)ContinuouslyandNormallyDistributedEndogenousVariablesNOTE:AtthistimeAMOSCANNOThandlenotcontinuouslydistributedoutcomevariablesSEMAssumptions(cont’d)ModelIdentificationPis#ofmeasuredvariables[P*(P+1)]/2Df=[P*(P+1)]/2-(#ofestimatedparameters)IfDF0modelisoveridentifiedIfDF=0modelisjustidentifiedIfDF0modelisunderidentifiedMissingdatainSEMTypesofmissingdataMCARMissingCompletelyatRandomMARMissingatRandomMNARMissingNotatRandomHandlingMissingdatainSEMListwisePairwiseMeansubstitutionRegressionmethodsExpectationMaximization(EM)approachFullInformationMaximumLikelihood(FIML)**Multipleimputation(MI)**Thetwobestmethods:FIMLandMISEMSoftwareSeveraldifferentpackagesexistEQS,LISREL,MPLUS,AMOS,SAS,...ProvidesimultaneouslyoveralltestsofmodelfitindividualparameterestimatetestsMaycomparesimultaneouslyRegressioncoefficientsMeansVariancesevenacrossmultiplebetween-subjectsgroupsIntroductiontoAMOSAMOSAdvantagesEasytouseforvisualSEM(StructuralEquationModeling).Easytomodify,viewthemodelPublication–qualitygraphicsAMOSComponentsAMOSGraphicsdrawSEMgraphsrunsSEMmodelsusinggraphsAMOSBasicrunsSEMmodelsusingsyntaxStartingAMOSGraphicsStartProgramsAmos5AmosGraphicsReadingDataintoAMOSFileDataFilesThefollowingdialogappears:ReadingDataintoAMOSClickonFileNametospecifythenameofthedatafileCurrentlyAMOSreadsthefollowingdatafileformats:AccessdBase3–5MicrosftExcel3,4,5,and97FoxPro2.0,2.5and2.6Lotuswk1,wk3,andwk4SPSS*.savfiles,versions7.0.2through13.0(bothrawdataandmatrixformats)ReadingDataintoAMOSExampleUSEDforthisworkshop:CondomuseandwhatpredictorsaffectitDATASET:AMOS_data_valid_condom.savDrawinginAMOSInAmosGraphics,amodelcanbespecifiedbydrawingadiagramonthescreen1.Todrawanobservedvariable,clickDiagramonthetopmenu,andclickDrawObserved.Movethecursortotheplacewhereyouwanttoplaceanobservedvariableandclickyourmouse.Dragtheboxinordertoadjustthesizeofthebox.Youcanalsouseinthetoolboxtodrawobservedvariables.2.Unobservedvariablescanbedrawnsimilarly.ClickDiagramandDrawUnobserved.Unobservedvariablesareshownascircles.Youmayalsouseinthetoolboxtodrawunobservedvariables.DrawinginAMOSTodrawapath,Click“Diagram”onthetopmenuandclick“DrawPath”.Insteadofusingthetopmenu,youmayusetheToolBoxbuttonstodrawarrows(and).DrawinginAMOSTodrawErrorTermtotheobservedandunobservedvariables.Use“UniqueVariable”buttonintheToolBox.Clickandthenclickaboxoracircletowhichyouwanttoadderrorsorauniquevariables.(WhenyouuseUniqueVariablebutton,thepathcoefficientwillbeautomaticallyconstrainedt