Joint Use of Mediation and Principal Stratification Approaches

• Activity Number: 103
• Type: Invited
• Date/Time: Monday, August 4, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Mental Health Statistics Section
• Author(s): Booil Jo*+ and Elizabeth A. Stuart
• Companies: Stanford University and Johns Hopkins Bloomberg School of Public Health
• Keywords: causal mediation ; principal stratification ; sensitivity analysis

Abstract

In dealing with entirely or partially unobservable quantities to do causal mediation modeling, we intentionally (because there are too many possibilities) and/or unintentionally (because our knowledge and imagination are limited) narrow the set of explicit assumptions that would make causal interpretation possible. With these fundamental difficulties, accurately defining causal mediation effects and conducting convincing sensitivity analysis are both unlikely to be perfect. Further, providing careful and thoughtful suggestions for future interventions is often difficult given the variety of possible explanations of the results from a particular mediation analysis. As one way of improving this situation, we propose to jointly utilize multiple approaches to investigating mediational processes. In particular, we will focus on the joint use of the causal mediation (e.g., Imai et al., 2010; VanderWeele, 2009, 2010) and principal stratification (e.g., Elliott et al., 2010; Jo, 2008; Jo et al., 2011) approaches. We will demonstrate that the joint use of the two approaches not only improves sensitivity analysis, but also enriches interpretation of mediational processes.