Conventional and Principal Stratification Perspectives on Mediation

• Activity Number: 326
• Type: Invited
• Date/Time: Tuesday, August 2, 2011 : 10:30 AM to 12:20 PM
• Sponsor: Social Statistics Section
• Author(s): Booil Jo*+ and Elizabeth A. Stuart
• Companies: Stanford University and The Johns Hopkins University
• Keywords: Baron and Kenny approach ; McArthur approach ; Principal stratification ; Causal inference ; Mediation ; Conditional ignorability

Abstract

We will first review how mediation is understood in conventional approaches such as the Baron/Kenny and the McArthur approach. This is an important step given that these approaches are widely used in many areas of social, behavioral, and medical research. Then, we will clarify how their underlying assumptions are interpreted in contemporary causal mediation approaches such as principal stratification. Principal stratification refers to cross-classification of individuals based on potential values of posttreatment variables under all compared treatment conditions. Since the resulting strata are unaffected by treatment, treatment effects conditioning on principal strata can be interpreted as causal effects. Finally, we will focus on the conditional ignorability (i.e., conditional on pretreatment covariates, individuals with different mediator status are comparable), which is the central assumption necessary for causal interpretation in conventional mediation analyses, and is often strongly criticized. We will examine implications of this assumption from the principal stratification perspective and compare the utility of the assumption in conventional and causal mediation modeling.