Marginal structural models for observational studies with time-dependent treatments
*Daniel Rubin, FDA 

Keywords: marginal structural models

This talk provides an overview of marginal structural models, which were proposed by Robins as a method for estimating treatment effects in observational studies when a subject’s treatment regimen can change over time. After surveying the model and the problem formulation, I will describe several applications, methodological extensions, and the potential utility in postmarket safety assessments.