Abstract: In this paper we consider the problem in causal inference of estimating the local complier average causal effect (CACE) parameter in the setting of a randomized clinical trial with a binary outcome, cross-over noncompliance, and unintentional missing data on the responses. We 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. If conclusions are insensitive over a range of logically possible values of the sensitivity parameters, then the number of interpretations of the data is reduced, and causal conclusions are more defensible. We illustrate our methods using a randomized encouragement design study on the effectiveness of an influenza vaccine.
Taylor, L and Zhou, Xiao-Hua, "Relaxing Latent Ignorability in the ITT Analysis of Randomized Studies with Missing Data and Noncompliance" (February 2009). UW Biostatistics Working Paper Series. Working Paper 257.