Statistical Power and Optimal Design of Cluster-Randomized Trials Investigating Mediation Effects

• In Event: New Insights for Designing Multilevel Models: Predicting Random Effects, Using Predictor Aggregates, and Mediation Planning

Abstract

Cluster-randomized trials (CRTs) investigating mediation effects can additionally investigate the potential causal pathways between interventions and outcomes, providing theoretical guidance for the iterative improvement of interventions. When designing CRTs investigating these mechanisms, sample size planning must consider (1) the sample size allocation and total clusters needed to achieve adequate statistical power for both mediation and total effects and (2) the cost efficiency of the study design. The present study has developed closed-form statistical power formulas for CRTs investigating mediation effects with and without covariates. We have further developed an optimal design framework to identify jointly optimal sample allocations targeting both effects that provide the maximum statistical power under a fixed budget.

Authors

• Zuchao Shen, University of Georgia

• Benjamin Kelcey, University of Cincinnati

• Wei Li, University of Florida

• Walter L. Leite, University of Florida