Sample size and power calculations for causal mediation analysis: A Tutorial and Shiny App.

Area of Science:

• Statistics
• Biostatistics
• Epidemiology

Background:

• Power analysis is essential for causal mediation studies.
• Existing methods for power analysis in causal mediation are underdeveloped.

Purpose of the Study:

• To propose a simulation-based power analysis method for regression-based causal mediation analysis.
• To provide an accessible web application for power and sample size calculations.
• To address the gap in power analysis tools for causal mediation.

Main Methods:

• A simulation-based approach using repeated sampling from predefined population models.
• Calculation of power based on the proportion of significant test results across replications.
• Utilizing Monte Carlo confidence intervals for efficient and accurate inference, compatible with the 'mediation' R package.

Main Results:

• The proposed method provides a faster alternative to bootstrapping for power analysis.
• The tool is versatile, applicable to various treatment, mediator, and outcome types (binary or continuous).
• Sample size recommendations are provided for different study scenarios.

Conclusions:

• The developed method and application facilitate robust study design in causal mediation analysis.
• Researchers can efficiently determine necessary sample sizes for adequate statistical power.
• The tool enhances the reliability of causal mediation effect detection.