心理学研究中的中介效应分析意义及方法评述

A review of the significance and methods of mediation analysis in psychological researches

中介效应分析有助于揭示心理现象间复杂的间接关系,也有助于间接分析横断研究中变量间的因果关系。常用的中介效应分析方法有依次检验法、Sobel方法、区间估计法,从统计功效和适用广度来看,区间估计法中的非参数百分位Bootstrap方法最为优秀。中介效应分析不仅要探明效应是否存在,还需要计算效应量大小,效应量指数有ab/c、k2和R2med三种,相比较而言,R2med在衡量效应量时较为稳定。中介效应分析还处于初级阶段,更复杂中介模型的理论架构及分析方法探讨将是未来中介效应研究的趋势。

Mediation effect analysis is conducive to reveal the indirect relationship among the mental phe-nomena and indirectly analyze the causal relationship among variables in cross-sectional study. The commonly usedmethods of mediation analysis include steps approach, Sobel method and interval estimation method. From statisticalpower and scope of application, nonparametric percentile bootstrap method is the best method in interval estimationmethods. Mediation analysis not only needs to prove whether the effect exists, but also needs to calculate the effectsize. The index of effect size refers to ab/c, k2 R2,. In comparison, R2 is more stable. Mediation analysis is still inthe initial stage now. The research about the more complicated theoretical framework and analysis method of media-tion model will be the trend in the future.