ISSN 0439-755X
CN 11-1911/B
主办:中国心理学会
   中国科学院心理研究所
出版:科学出版社

心理学报 ›› 2006, Vol. 38 ›› Issue (03): 448-452.

• • 上一篇    下一篇

有中介的调节变量和有调节的中介变量

温忠麟;张雷;侯杰泰   

  1. 华南师范大学心理应用研究中心,广州 510631

  • 收稿日期:2005-06-29 修回日期:1900-01-01 出版日期:2006-05-30 发布日期:2006-05-30
  • 通讯作者: 温忠麟

Mediated Moderator and Moderated Mediator

Wen-Zhonglin,Zhang Lei,Hou Jietai   

  1. Faculty of Educational Science, South China Normal University, Guangzhou 510631, China
  • Received:2005-06-29 Revised:1900-01-01 Published:2006-05-30 Online:2006-05-30
  • Contact: Wen Zhonglin

摘要: 研究了有中介的调节效应模型和有调节的中介效应模型,两种模型都涉及到中介变量和调节变量。分别讨论了这两种模型的分析方法,两者基本上都是按中介效应模型的分析步骤,但其中都有一步要做调节效应的分析。还讨论了同时包含有中介的调节变量和有调节的中介变量的混合模型。最后,作为示范例子,用混合模型分析儿童行为对同伴关系的影响,结果发现,如果混合模型中的调节都设定是线性的,那么调节变量对因果关系是二次调节

关键词: 调节效应, 中介效应, 有中介的调节变量, 有调节的中介变量, 混合模型

Abstract: We presented two causal models that include either a mediated moderator or a moderated mediator. The “mediated moderator” model focuses on the moderating effect and the “moderated mediator” model focuses on the mediating effect. Both models can be analyzed using procedures similar to those for testing mediating effects. Each procedure, however, also consists of an extra step to test moderating effects. These procedures are summarized below:
The moderating effect of U on the relation between Y and X is mediated by W:
(1) Regress Y on , and . A significant coefficient associated with UX implies that U is the moderator of the relation between Y and X.
(2) Regress W on , and UX and look for a significant coefficient associated with UX.
(3) Regress on , , and W. A significant coefficient associated with W suggests the mediating effect of W on the moderated (by U) relation between X and Y. Moreover, the moderated effect of U on the relation between Y and X is completely mediated by W if the coefficient associated with is not significant.
The mediating effect of W on the relation between Y and X is moderated by U:
(1) Regress Y on , and look for a significant coefficient associated with X;
(2) Regress W on , and look for a significant coefficient associated with X;
(3) Regress on , and W, and look for a significant coefficient associated with W;
Steps 1-3 are used to test the mediating effect of W on the relation between Y and X .
(4) Regress on , , W and UW. A significant coefficients associated with UW suggests that the mediating effect of W on the relation between X and Y is moderated by U.
In addition to these two models, we also presented a mixed model including both a mediated moderator and a moderated mediator. Using an example to illustrate the mixed model, we showed that Teacher Liking as a mediator on the relation between students’ Antisocial Behavior and Peer Acceptance was moderated by Perceived Authoritative Teachering, and that Perceived Authoritative Teachering as a moderator on the relation between Antisocial Behavior and Peer Acceptance was mediated by Teacher Liking. The relative size of the mediating effect in the mixed model was also computed. It was further demonstrated that the moderator of the mediated relation represented a quadratic effect, whereas other relations examined in the mixed model were linear effects

Key words: moderating effect, mediating effect, mediated moderator, moderated mediator mixed model

中图分类号: