A Multiple Mediator Model: Power Analysis Based on Monte Carlo Simulation

Abstract

Most of the applied psychological researchers usually conduct studies requiring application of advanced mediation models, such as multiple mediator models. However, in designing research, most of the applied researchers largely ignore the statistical power of their studies. As a result, power analyses are ignored when researchers report their results. It is well recognized that low power is one possible reason for no statistically significant result being identified in a study. Moreover, studies with low statistical power have been labeled “scientifically useless”. The current study describes how to apply Monte Carlo simulation to test the type I error rates and statistical power of mediating effects in a multiple mediator model. Findings from the current simulation study indicated that the effect sizes of mediating effects and sample sizes were two important factors influencing type I error rates of indirect effects in a multiple mediator model. Furthermore, the requirement of sample size and desired power level were strongly depended on the effect size of the indirect effect.