香港中文大学 基于lisrel的SEM讲义Note3

1PathAnalysisandStructuralModel结构模型和全模型张伟雄博士香港中文大学工商管理学院副院长管理学系教授GordonW.Cheung,Ph.DProfessor,DepartmentofManagementAssociateDean,FacultyofBusinessAdministrationTheChineseUniversityofHongKong2PathModel结构模型X1X2X3Y1Y2Y311121323323121322133ExogenousandEndogenousVariables•ExogenousVariable(外源变值)actsonlyasapredictoror“cause”forothervariablesinthemodel.Inpathdiagrams,theexogenousvariableshaveonlycausalarrowsleadingoutofthemandarenotpredictedbyanyothervariablesinthemodel.•Endogenousvariable(内生变值)isthedependentoroutcomevariableinatleastonecausalrelationship.Intermsofapathdiagram,thereareoneormorearrowsleadingintotheendogenousvariable.Hair,etal(1995).MultivariateDataAnalysiswithReadings.Pp.619PrenticeHall.4PathAnalysis•Directeffect(直接效应)isthatinfluenceofonevariableonanotherthatisunmediatedbyanyothervariablesinapathmodel•Indirecteffects(间接效应)ofavariablearemediatedbyatleastoneinterveningvariable•Totaleffects(总效应)arethesumofdirectandindirecteffects•Totaleffects=Directeffect+IndirecteffectsBollen(1989).StructuralEquationswithLatentVariables.Pp.36.Wiley.5PathAnalysis:Example1•Bollen(1989)StructuralEquationswithLatentVariables.pp.82.Wiley.•Thevariablesareyearsintextilemill(x1),age(x2)•deference(orsubmissiveness)tomanagers(y1),supportforlaboractivism(y2)andsentimenttowardunions(y3)•Samplesize=1736PathAnalysis:Example1•LISRELProgram:Path1.ls8X1X2Y1Y2Y31222313121322137PathAnalysis:Example2•Netemeyer,R.,Burton,S.,&Johnston,M.(1995).Anestedcomparisonoffourmodelsoftheconsequencesofroleperceptionvariables.OrganizationalBehaviorandHumanDecisionProcesses,61,77-93.•Roleconflict(RC);RoleAmbiguity(RA);RoleOverload(RO)•Tension(T);JobSatisfaction(JS);OrganizationCommitment(OC);LeaveIntention(LI);Turnover(TO)•SampleSize=1818PathAnalysisExample2:Model1322122RCTRA111231213243JSLITO129PathAnalysisExample2:Model212RCRA111242T2132JSOC43LI54TO22322110PathAnalysisExample2:Model3131312RCRARO1211JS213243OCLITO2311PathAnalysisExample2:Model422RCJSRA111231213243OCLITO1212PathAnalysisExample2:InputFiles•Model1:PATH2_1.ls8•Model2:PATH2_2.ls8MIofTE(2,2)=11.88•Model2b:PATH2_2b.ls8•Model3:PATH2_3.ls8MIofTE(1,1)=20.03•Model3b:PATH2_3b.ls8MIofBE(3,1)=11.56•Model3c:PATH2_3c.ls8•Model4:PATH2_4.ls8MIofTE(1,1)=11.54•Model4b:PATH2_4b.ls813PathAnalysisExample2:ResultsModelChi-squaredfRMSEACFINNFISRMRECVI13.801850.01.01.01080.03140.2416218.5620100.06460.98540.96930.06220.37882b5.260290.01.01.01490.02780.3146338.6471120.10120.95720.92500.10340.45053b16.5488110.05090.99110.98300.05190.36183c4.4312100.01.01.01880.01960.2556415.572570.07850.98270.96300.05620.30724b2.759060.01.01.01630.01870.236014Example2:SquaredMultipleCorrelationforReducedFormModelTensionSatisfactionCommitmentLeaveIntentTurnover10.19700.27020.25910.02962b0.19700.44640.32930.23180.02653c0.46170.32410.23610.02704b0.44630.32890.22970.0262Model4b:PATH2_4b.ls815FullModel1X1X2X3X41234413121111y1y2y3y41234413121112y5y6y7y856788272625211212116FullStructuralEquationModel•MeasurementModelisusedtostudythemeasurementpropertiesofobservedvariables•PathModelisusedtoestimate“causal”relationshipsamongdirectlyobservedvariables•FullStructuralEquationModelcombinesmeasurementmodelandpathmodel17FullModelExample1•LISRELExample5.6:ModifiedModelforPerformanceandSatisfaction•Bagozzi,R.P.(1980).Performanceandsatisfactioninanindustrialsalesforce:Anexaminationoftheirantecedentsandsimultaneity.JournalofMarketing,44,65-77–Performance(PERF)MeasuredwithP1–JobSatisfaction(JS)MeasuredwithJS1&JS2–AchievementMotivation(AM)MeasuredwithAM1&AM2–Task-SpecificSelfEsteem(SE)MeasuredwithSE1&SE2–VerbalIntelligence(VI)MeasuredwithVI1•SampleSize=122•PathDiagram:Full1.GIF•InputFile:Full1.ls818Homework2:FullModelExample2•Pedhazur(1991).Measurement,DesignandAnalysis:AnIntegratedApproach.pp.727LawrenceErlbaum.•AssumethatAcademicAchievement(AA),MentalAbility(MA)andSelf-Concept(SC)wereeachmeasuredby3items.•ItishypothesizedthatMAaffectsbothAAandSC,whereasAAaffectsSC.Datawerecollectedfrom200students.19Homework2:KMandSDLAAA1AA2AA3SC1SC2SC3MA1MA2MA3KM1.521.45.581.29.34.231.31.30.28.531.27.36.35.48.451.38.35.46.41.38.421.42.44.48.39.31.41.691.37.39.43.38.36.40.77.731SD4.675.815.122.763.122.9310.1211.0912.3120Homework2:FullModelExample2•InputFile:Full2.ls8