中介效应分析:方法和模型发展

Analyses of Mediating Effects: The Development of Methods and Models

在心理学和其他社科研究领域,大量实证文章建立中介效应模型,以分析自变量对因变量的影响过程和作用机制。检验中介效应最流行的方法是Baron和Kenny的逐步法,但近年来不断受到批评和质疑,有人甚至呼吁停止使用其中的依次检验,改用目前普遍认为比较好的Bootstrap法直接检验系数乘积。本文对相关的议题做了辨析,并讨论了中介分析中建立因果关系的方法。综合新近的研究成果,总结出一个中介效应分析流程,并分别给出显变量和潜变量Mplus程序。最后介绍了中介效应模型的发展。

Mediation models are frequently used in the research of psychology and other social science disciplines. Mediation indicates that the effect of an independent variable on a dependent variable is transmitted through a third variable, which is called mediator. In most applied research, Baron and Kenny＇s （1986） causal steps approach has been used to test mediating effect. In recent years, however, many methodological researchers questioned the rationality of the causal steps approach, and some of them even attempted to stop its use. Firstly, we clarify the queries on the causal steps approach one by one. Secondly, we propose a new procedure to analyze mediating effects. The new procedure is better than any single method that constitutes the procedure in terms of Type I error rate and power. The proposed procedure can be conducted by using observed variables and/or latent variables. Mplus programs are supplied for the procedure with observed variables and/or latent variables. Finally, this article introduces the development of mediation models, such as mediation model of ordinal variables, multilevel mediation, multiple mediation, moderated mediation, and mediated moderation.