Abstract

Mediation analysis is a highly useful tool for psychological researchers seeking to identify a third variable that explains the relationship between two variables. To enhance the interpretive validity of mediation effects, researchers must clearly understand the assumptions of mediation analysis and the characteristics of their data. This study identified issues in the recent reporting practices of mediation analysis results in psychological research and proposed factors that could influence these results. Furthermore, we have proposed a set of recommendations for researchers who employ mediation analysis in presenting their findings. This study employed a computer simulation method and demonstrated that various alternative explanations were correct. Specifically, nine simulations were designed to investigate the frequency of false mediation effect detection or whether the mediation effect is detected when the assumptions of mediation and regression analyses were not met or when there is no meaningful mediation effect. In light of these findings, the study proposed a set of guidelines to aid in the reporting of results. In light of these findings, the present study has put forth the recommendations: 1) demonstrating the logical validity of temporal sequence and comparing the equivalent model with the reversed temporal sequence to the research model; 2) conducting sensitivity analysis; 3) using valid measurements; 4) checking data prior to analysis; and 5) providing information for assessing the possibility of Type I errors.