Causal Mediation Programs in R, Mplus, SAS, SPSS, and Stata.

Area of Science:

• Statistics
• Biostatistics
• Epidemiology

Background:

• Mediation analysis investigates the pathway through which an independent variable (X) affects an outcome (Y) via a mediator (M).
• Causal mediation methods, grounded in potential and counterfactual outcomes frameworks, offer a robust approach to understanding causal pathways.
• Existing software programs for estimating causal mediation effects exhibit considerable variation in setup, estimation techniques, output, and platform.

Purpose of the Study:

• To compare the performance and output of different software programs designed for causal mediation analysis.
• To evaluate the consistency of causal effect estimates across various programs using a standardized empirical example.

Main Methods:

• An empirical example was utilized to estimate a single mediator model with an X-M interaction.
• The model featured a continuous mediator and a continuous outcome.
• The same model was implemented and analyzed across multiple available software packages for causal mediation analysis.

Main Results:

• Despite employing diverse estimation methodologies, the evaluated software packages produced comparable causal effect estimates.
• Estimates were consistent for mediation models involving a continuous mediator and a continuous outcome.
• The study identified similarities and unique features across the compared programs.

Conclusions:

• Current software programs are largely consistent in estimating causal mediation effects for continuous mediator-outcome models.
• Researchers can be confident in the generalizability of results across different platforms for these specific model types.
• The paper provides a detailed comparison and recommendations to guide software selection in causal mediation analysis.