中介效应检验程序及其应用-温忠麟

:2004-02-09*00(DBA010169)():,E-mail:[e-mail]*温忠麟1,2张雷2侯杰泰2刘红云3(1,510631)(2,)(3,100875);;,Sobel,,,,,,,B841.21(mediator),,PsycINFO,1998~2003(mediatormediatingmediation)1100,/0/0,1998~2003/0/020,,;,,[1][2],,[1],/0,,(),,,,MacKinnon[3]1998(power)[4],,[3],MacKinnon[4],,/0()2,(Y)(X),,/XY0/0,,(5.1)2.1XY,XMY,M,/0/0,/0[5],/2004,36(5):614~620ActaPsychologicaSinica6140()/0()/0[6],/0/0(),(1):Y=cX+e1(1)M=aX+e2(2)Y=c.X+bM+e3(3)YX,c(H0:c=0),MM,(mediatoreffect)?[3]1[7,8],:(i);(ii),(),BaronKenny()[8]:(iii),,JuddKenny[7],(1):(i)c(H0:c=0);(ii)a(H0:a=0),b(H0:b=0):(iii)c.ab,H0:ab=0,,[4,9,10],abc.c,H0:c-c.=0,,[11,12]2.2[13],(indirecteffect)1,cXY,abM(),c.,c=c.+ab(4),(4)[13](4)([14])(4)c-c.=ab,c-c.,H0:ab=0H0:c-c.=0,,,,,,(),,,,,,([15],p128;[13],p48)Y,X,M(X)(M)0.707(=a),(M)(Y)0.707(=b),-0.50(=cc)(=c)()()(suppression),,,,,,3,,([13],[16])(1)5:615,(recursive),,,,,(1))(3),,,,c^c;a,b,cca^,^b,^cc,a^^bc^c-^cc,,[14],a^^b,,(a^^b/(^c.+a^^b)),(a^^b/^c.),[14],3.1,,H0:a=0H0:b=0,,H0:c.=0tt,,,,,a(),b(),,ab(),(H0:a=0b=0,ab),[3],3.2H0:ab=0H0:ab=0a^^b5[3](500),SobelTaylor[9,10]sab=a^2s2b+^b2s2a(5),sa,sba^,^bz=a^^b/sab,LIS-REL[17]/OU0/EF0,t,tSobelz,XMY,,z=a^^b/sabMacKinnon[4],0.050.97(1.96,,,),~davidpm/ripl/methods.htmMacKinnon[3],,,a=0b=00.05,,,a=0b=0(ab=0,),0.05,3.3H0:c-c.=0,H0:c-c.=0^c-^c.MacKin-non[3],a=0b=00.05Clogg[11]sc-c.=šrXMšsc.(6)rXMXMFreed-man[12]sc-c.=s2c+s2c.-2scsc.1-r2XM(7)a=0bX0(ab=0,),(t=(^c-^c.)/sc-c.)(6),100%,(6)H0:b=0[11],(ab=0),b,()616363.4(),::(1)a=b=0,(2)a=0,(3)b=0;abX0,1,MacKinnon[3]1a=b=0a=0b=0abX0JuddKennytN-2=a^/sa,tN-3=^b/sb,tN-3=c^/sc.[7]n0.05F0.05F0.05BaronKenny[8]tN-2=^a/sa,tN-3=^b/sb,n0.050.050.05H0:ab=0Sobel[9]z=a^^b/a^2s2b+^b2s2an0.050.050.05MacKinnon[4],0.05H0:c-c.=0Clogg[11]tN-3=(^c-^c.)/|rXMsc.|0.050.05Freedman[12]tN2=^c-^c.s2c+s2c.-2scsc.1-r2XM0.050.05:N0n0.0500.05c.=0,a=0b,N;b=0a,N;a,b,N3.5,,,,1,1.c,,22.BaronKenny[8],a,b,,XYM,0.05,3,(),,43.JuddKenny[7](),c.,,,XYM;,,XYM4.Sobel[9],,M,2,Sobel,(5)sabz=a^^b/sabLISREL,24(X)(Y)5:617,[18,19],(X),9(),(Y),,()(U),9(,,)(W)(/0/05)N=595,,(),:Y18.87W1.130.45X-9.78-2.2094.25U0.630.09-0.220.56,,4.1,,(W)2,,(3t),t,,01338@01349/01232=5018%2(W)y=-01232xSE=0.040,t=-5.8**w=-01338xSE=0.039,t=-8.7**y=0.349wSE=0.040,t=8.7**-0.114xSE=0.040,t=-2.8**:SE**p0.01W:,,,,,,,,4.2,,(U)()3(ux,t=-0.72,p0.05),,Sobel,z=a^^b/a^2s2b+^b2s2a,a^=-0.030,sa=0.041,^b=0.187,sb=0.039z=-0.72,p0.05(U)3(U)y=-01232xSE=01040,t=-5181**u=-01030xSE=01041,t=-0172y=01187uSE=01039,t=4179**-01226xSE=01039,t=-5179**:SE**01015511,,(),/0,/0X1X2M1M2,[20](X1X2M1M2)(),(1))(3),:Y=c1X1+c2X2+c3X1X2+e1M1=a11X1+a12X2+a13X1X2+e12M2=a21X1+a22X2+a23X1X2+e22M1M2=a31X1+a32X2+a33X1X2+e3261836Y=c.1X1+c.2X2+c.3X1X2+b1M1+b2M2+b3M1M2+e3,,(2)c,a,b,c.,(c.),X1M2,c,a,b,c.c1,a21,b2,c.1;X1X2M1,c,a,b,c.c3,a13,b1,c.3;X2M1M2,c,a,b,c.c2,a32,b3,c.2,M1M2,,c,a,b,c.(),,(4),X1M2,c,a,b,c.c1,a21,b2,c.1c1=c.1+a21b25.2,,,,,,(),,,,,1ShenJ,WangD,PengH,TangD.Theeffectsofmediatorsontheagingofprimarymentalabilities(inChinese).ActaPsycholog-icaSinica,2003,35(6):802~809(,,,..,2003,35(6):802~809)2FangP,XiongD,GuoC.Theeffectofparentingstylesonchi-ldren.sacademicachievement(inChinese).PsychologicalSc-ience,2003,(1):78~81(,,..,2003,(1):78~81)3MacKinnonDP,LockwoodCM,HoffmanJM,WestSG,SheetsV.AComparisonofmethodstotestmediationandotherin-terveningvariableeffects.PsychologicalMethods,2002,7(1):83~1044MacKinnonDP,LockwoodCM,HoffmanJM.Anewmethodtotestformediation.PaperpresentedattheannualmeetingoftheSocietyforPreventionResearch,ParkCity,UT.1998,June5DuncanOD,FeathermanDL,DuncanB.Socioeconomicback-groundandachievement.NewYork:SeminarPress,19726JamesLR,BrettJM.Mediators,moderatorsandtestsformedia-tion.JournalofAppliedPsychology,1984,69(2):307~3217JuddCM,KennyDA.Processanalysis:Estimatingmediationintreatmentevaluations.EvaluationReview,1981,5(5):602~6198BaronRM,KennyDA.Themoderator-mediatorvariabledistinc-tioninsocialpsychologicalresearch:Conceptual,strategic,andstatisticalconsiderations.JournalofPersonalityandSocialPsycho-logy,1986,51(6):1173~11829SobelME.Asymptoticconfidenceintervalsforindirecteffectsinstructuralequationmodels.In:SLeinhardt(Ed.).Sociologicalmethodology1982.Washington,DC:AmericanSociologicalAsso-ciation,1982.290~31210SobelME.Directandindirecteffectsinlinearstructuralequationmodels.In:JSLong(Ed.)Commonproblems/propersolutions.BeverlyHills,CA:Sage,1988.46~6411CloggCC,PetkovaE,ShihadehES.Statisticalmethodsforana-lyzingcollapsibilityinregressionmodels.JournalofEducationalStatistics,1992,17(1):51~7412FreedmanLS,SchatzkinA.Samplesizeforstudyingintermediateendpointswithininterventiontrialsofobservationalstudies.Amer-icanJournalofEpidemiology,1992,136:1148~115913BollenKA.Structuralequationswithlatentvariables.NewYork:Wiley,198914MacKinnonDP,WarsiG,DwyerJH.Asimulationstudyofme-diatedeffectmeasures.MultivariateBehavioralResearch,1995,30(1):41~6215McfatterRM.Theuseofstructuralequationmodelsininterpret-ingregressionequationsincludingsuppressorandenhancervar-iables.AppliedPsychologicalMeasurement,1979,3:123~13516HauKT,WenZ,ChengZ.StructuralEq