Sensitivity Analysis under Unmeasured Confounding

Please join the GW Department of Statistics for a seminar with Zhiqiang Tan, PhD, distinguished professor at the Rutgers University Department of Statistics.

Abstract: 

For point identification of average treatment effects, the assumption of no unmeasured confounding asserts that the treatment is conditionally independent of potential outcomes among individuals with the same measured covariates as in a stratified randomized trial. This assumption is statistically untestable and may in general be violated in observational studies. A sensitivity analysis investigates how the treatment effects might vary under possible unmeasured confounding. In this talk, I will discuss marginal sensitivity models, related theory and methods, and an application to an observational study on the effects of right heart catheterization.

About the Speaker: 

Zhiqiang Tan's research interests include causal inference, Monte Carlo methods, statistical learning and related areas. He is a Fellow of the American Statistical Association and the Institute of Mathematical Statistics and an elected member of the International Statistical Institute.