1、 2006, 38(3):448 452 Acta Psychologica Sinica448 l :2005 - 06 - 29* Sv 51 “(05JZD00034)、 S S“ E”? 5(DBA010169)#Pv。YT:, E -m ail:w enzl scnu. edu. cnM M *温忠麟1, 2 张 雷3 侯杰泰3(1 2 =Sv ,。 ,U =3.9581. 297,XY r, U9F79F,U HrKv, /,U =6. 260r。 , 98%Ul1. 497(M (2S),_ V?C,U1. 2971. 497W H,XY()rA。5 ) S 5, yT “ 1M
2、 yM , V ? #B M ,t M V ? M ,9 V ?M 。DnVBt HcBM , HcBM 1 n,7 HcM M Fhn。 HctM 7y,7 Ad9 f /, 、 Y。B, yT msT =_ B“ ( “ ) A 。 ,1syT1“( YX1“), M M , VYV mM W1“T %s。 s Q:M M 。?C, T L,9 L, * M yT1“ =Q。 r4 Ad9 d。QV , T =Q , V I n| B , Hc M M 。 #M AM , ) ms .M “ a。 .M , V| mZ (VAM ) (V .M )。“ V In c .M TM 、M 、M
3、 、M f。V, T .M , MF。y .M 1 S, “ BZ(Z), Z。1 Z , E_M1 。7AM f, E Yz, Vrs。B4 ,M , ? PYS。erBZ :Y =b0 +b1X 1 +b2X2 +b3X1X 2 +eT YSE,|Y,X1,X 2, X1X 2X1X 2sYSMZY , ZX 1, ZX 2ZX 1X 2, pS, 。“S, b1, b2b3sY ZX 1, ZX 2ZX 1X 2B“ ,ZX 1X 2 ZX 1ZX 2( T),y7b3Vr。 07, P V ZE , A K。$ k 1), 、)vl、) r, 3 k H V ? s。4 E 1“(Y)
4、 , TB 3 )v, $4 v。 ABf,T , )30 H, s 30( 9 V4 1-), )40 H, s 40。 h“)vlY,BSE |M )SS(nD 5 )。srR PS ,“F ”) h“)vlY。“B) H,)y V ?YM ,Y V ? eYV ) Vh“。V, 5 ) 。 ID1 Wen Z, H auK T, C hang L. A C omparison ofM oderator andMedia-tor and TheirApplications(in Chinese). Acta Psycho logica Sinica,2005, 37(2):268 274(
5、, ,f .rr1 . , 2005, 37(2):268 274)2 Wen Z, Chang L, H au K T, L iuH. Testing andApplication of theM ediating E ffects (in Chinese). Acta Psychological Sinica, 2004,36(5):614 620(,f , , .r_# . , 2004, 36(5):614 620)3 Baron RM, K enny D A. Them oderator -mediator variable distinc-tion in social psycho
6、log ica l research:C onceptual, strategic, and sta-452 38 tistical considerations. Journal of Persona lity and SocialP sychology,1986, 51(6):1173 11824 Jam es L R, B rett JM. M ediators, m oderators and tests for m edia-tion. Journal of Applied P sychology, 1984, 69(2):307 3215 Chang L, Liu H, Wen Z
7、, Fung K, W ang Y, Xu Y. M ediatingteacher liking and m oderating authoritative teachering on C hineseado lescentsperceptions of antisocial and prosocial behaviors. Jour-nal of Educationa lPsychology, 2004, 96(2):369 3806 Bollen K A. Structura l equationsw ith latent variables. New York:W iley, 1989
8、7 H au K T, W en Z, Cheng Z. StructuralEquationM odel and ItsAp-plications (in Chinese). Beijing:Educational Science Press, 2004( ,0 .Z # .:S, 2004)8 M acK innon D P, W arsiG, Dwyer JH. A sim ulation study ofm edi-ated effect m easures. M ultivariate Behavioral Research, 1995, 30(1):41 62M ediatedM
9、oderator andM oderatedM ediatorW en Zhonglin1, 2, Chang Lei3 , Hau K it-Tai3(1 Faculty ofEducationalScience, South China NormalUniversity, Guangzhou 510631, China)(2 Hong Kong Examinations and Assessment Authority, HongKong, China)(3Department of EducationalPsychology, TheChineseUniversity ofHong Ko
10、ng, Hong Kong, China)AbstractW e presented two causalm odels that include either am ediated moderator or a moderated m ediator. The “m ediatedm oderator“model focuses on the moderating effect and the “moderated m ediator“model focuses on themediating effec.tBoth m odels can be analyzed using procedu
11、res sim ilar to those for testingmediating effects. Each procedure, however, al-so consists of an extra step to testm oderating effects. These procedures are summ arized below:Them oderating effect ofU on the relation betweenY and X is mediated byW:(1)RegressY onX, U and UX. A significant coefficien
12、t associated w ithUX implies thatU is them oderator of the re-lation betw eenY and X.(2)RegressW on X,U andUX and look for a significant coefficient associated w ithUX.(3)RegressY on X,U, UX andW. A significant coefficient associated w ithW suggests themediating effect ofW onthe moderated (byU) rela
13、tion between X andY. Moreover, themoderated effect ofU on the relation betweenY andX iscompletely mediated byW if the coefficient associated w ith UX is not significan.tThem ediating effect ofW on the relation betw een Y andX ismoderated byU:(1)RegressY on X,U and look for a significant coefficient
14、associated w ith X;(2)RegressW on X,U and look for a significant coefficient associated w ith X;(3)RegressY onX,U andW, and look for a significant coefficient associated w ithW;(Steps 1 -3 are used to testthe mediating effect ofW on the relation between Y and X).(4)RegressY onX,U, W and UW. A signif
15、icant coefficients associated w ithUW suggests that themediating effectofW on the relation between X and Y ism oderated byU.In addition to these twom odels, we also presented am ixed model including both am ediatedm oderator and amodera-ted m ediator. Using an example to illustrate them ixed model,
16、we showed thatTeacher Liking as am ediator on the relationbetween studentsAntisocialBehavior and PeerAcceptancewasm oderated by Perceived Authoritative Teachering, and thatPerceived Authoritative Teachering as am oderator on the relation between Antisocial Behavior and Peer Acceptance w asm ediated
17、by Teacher Liking. The relative size of themediating effect in them ixedm odelwas also computed. Itw as furtherdem onstrated that them oderator of the mediated relation represented a quadratic effect, whereas other relations exam inedin the m ixed m odelwere linear effects.K eywords moderating effect, m ediating effect, mediated m oderator, moderated mediator, m ixed m ode.l
