中介效应的检验方法和效果量测量:回顾与展望

通过中介效应检验方法之间的比较和效果量指标之间的比较,建议放弃将总效应c显著作为中介效应检验的前提条件,放弃基于直接效应c'显著性的完全和部分中介的提法,推荐使用偏差校正的百分位Bootstrap法直接对中介效应ab进行检验,使用κ²、R_med²等中介效果量指标并报告效果量的置信区间。作为示例,用R软件的MBESS软件包对某消防员饮食健康调查进行了中介效应检验和效果量测量。随后展望了中介效应检验方法和效果量测量的拓展方向。

Mediation Analysis and Effect Size Measurement:Retrospect and Prospect

Synthesizing and comparing the methods of mediation analysis and effect size index,we suggest the focus on the significance between the independent and dependent variables,both before and after mediation tests,is unjustified and can impair theory development and testing.We recommend that,1) mediator effect can be analyzed by bias-corrected percentile bootstrap method;2) full and partial mediation should be replaced byκ2,R2medindex,whenever possible,confidence intervals for the population effect size should be reported.Mediation analysis and effect size measurement are applied to a real-world example using data from a health promotion program to improve the nutrition habits of firefighters.MBESS package of R software is also provided for researchers to implement mediation analysis and effect size measurement.The future research direction of mediation analysis and effect size measurement is discussed at the end of the paper.