Monte Carlo Confidence Intervals for Specific Indirect Effects in Mediation Analysis

In mediation analysis, there are occasions where the causal chain has more than two mediators, which is termed a micromediational chain. The extension and application of methods to test and build confidence intervals (CIs) for an indirect effect in a micromediational chain with more than two mediators are necessary. We extended the application of the Monte Carlo method to build CIs for indirect effects in micromediational chains. We also implemented the Monte Carlo method in the RMediation package. Finally, we conducted a simulation study comparing the Type I error rates and power of the Monte Carlo CIs with percentile bootstrap and asymptotic normal distribution with multivariate delta standard error (Asymptotic–Delta) CIs. The results indicated that the Monte Carlo and percentile bootstrap methods performed similarly while both methods, in general, outperformed the Asymptotic–Delta method in terms of the Type I error rate and power.