Stepped Wedge Design for Multiple Interventions

Abstract

This thesis includes a brief summary of statistical models used in stepped wedge design (Chapter 1) and an introduction to variants for the stepped wedge design with multiple interventions (Chapter 2), including Concurrent, Replacement, Supplementation, and Factorial designs. Under the basic model with random cluster effect, nonparametric fixed time effects, and fixed treatment effect, estimates of the treatment effects and variances of the estimates are presented and compared. In some specific settings, the Supplementation design is found to be preferable to the Replacement design, and provides more precise estimates. In Chapter 3, the model with random time effects is discussed. Performances of different methods, including generalized estimating equation and linear mixed models, are presented in terms of the mean and standard deviation of the estimates of the treatment effect, coverage of the 95\% confidence intervals for the estimates, and mean of the estimates of the slope. Some limitations of this work are discussed in Chapter 4.