Leslie Taylor
Randomized clinical trials are a powerful tool for investigating causal treatment effects, but in human trials there are oftentimes problems of noncompliance and missing data which standard analyses, such as intention-to-treat, either ignore or incorporate in such a way that the resulting estimand is no longer a causal effect. One alternative to these analyses utilizes the principal stratification framework (Frangakis and Rubin, 2002) where we focus on the complier average causal effect (CACE) which estimates the average causal treatment effect among a subpopulation that would comply under any treatment assigned. In the first part, we estimate the CACE for a binary outcome and focus on the development of a moment estimator that relaxes the assumption of latent ignorability and incorporates sensitivity parameters that represent the relationship between potential outcomes and associated potential response indicators. In the second part, we develop estimators for the CACE using multiple imputation methods, which have been successfully applied to a wide variety of missing data problems, but have not yet been applied to the potential outcomes setting of causal inference. In the third part, we extend the multiple imputation methods to a clustered encouragement design study exploring the role of computer-based care suggestions in managing patients with chronic heart failure.
