Informative Priors for Unmeasured Confounding

• Activity Number: 455
• Type: Invited
• Date/Time: Wednesday, August 7, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Joe Hogan*+
• Companies: Brown University

Abstract

The presence of unmeasured confounding is more the exception than the rule in causal inference from observational data. Using data from a large electronic health records system from Kenya, we formulate models for the causal effect of a point exposure in terms of one or more parameters that represents the degree of unmeasured confounding. We show how to formulate priors for these parameters, and draw approximate posterior inference about treatment effects. The model structure builds on sensitivity analysis methods proposed by Robins (1997, Synthese).