A Simulation Study of Statistical Approaches to Data Analysis in the Stepped Wedge Design

Abstract

This paper studies model-based and permutation-based approaches to analyze data in the stepped wedge design under 9 scenarios. We compare robustness, efficiency, Type I error rate under null conditions, and power under alternative conditions for GEE and LMM based approaches. We find that GEE models with exchangeable correlation structures are more efficient than GEE models with independent correlation structures under all scenarios. The model-based GEE Type I error rate can be inflated when applied with a small number of clusters, but this problem is solved under a permutation approach. Correct model specification is more important to LMM as compared to GEE. However, in contrast to the model-based LMM results, the permutation-based Type I error rates for LMM models under scenarios with a random treatment effect has an unexpected inflation even though the models perfectly match the corresponding scenarios.