Testing for mediation effects under non-normality and heteroscedasticity: a comparison of classic and modern methods

Analysis of mediation effects is common in educational research. The most popular methods for testing mediation effects are the causal steps approach proposed by Baron and Kenny (1986), and Sobel's test (1982). Numerous studies have pointed out limitations of these classic procedures, and modern methods based on bootstrap techniques have been proposed. However, the robustness of these modern methods has not been thoroughly understood. In this study, we evaluate the impact of non-normality and heteroscedasticity on the performance of both classic and modern methods for mediation analysis. Results from our simulation show that non-normality and heteroscedasticity can substantially undermine the statistical power of all methods. Nevertheless, modern bootstrap techniques are preferred over classic methods.