Maximum Likelihood Estimation and Design and Inference Considerations for Sequential Multiple Assignment Randomized Trials

This chapter provides a discussion of the sequential multiple assignment randomized trial (SMART) design, its operational characteristics and advantages compared to alternative design options, and details on the statistical analyses associated with SMARTs, with motivating examples. Inference and sample size calculation in the SMART setting are addressed, focusing on the maximum likelihood estimation (MLE) approach, which helps to reveal some unique properties in statistical inference in SMARTs and is generalizable to other semi-parametric estimation methods. The chapter ends with final remarks about the benefits of our proposed methods and closing comments on the utility of SMARTs.