Proximal Causal Inference
Skepticism about the assumption of no unmeasured confounding -also known as exchangeability-, is often warranted in making causal inferences from observational data, because exchangeability hinges on an investigator's ability to accurately measure covariates that capture all potential sources of confounding. In practice, the most one can hope for is that covariate measurements are at best proxies of the true confounding mechanism, thus invalidating inferences made under exchangeability. In this talk, we consider the framework of proximal causal inference introduced by Tchetgen Tchetgen et al (2020), which while explicitly acknowledging covariate measurements as imperfect proxies of confounding mechanisms, offers an opportunity to learn about causal effects even if exchangeability on basis of measured covariates fails. We provide an overview of nonparametric identification conditions, as well as parametric, semi-parametric and nonparametric modes of inference in a variety of causal settings, including total effects, mediation analysis, interference on a network subject to homophily bias, Synthetic controls, and longitudinal analysis of time-varying treatment subject to time-varying unmeasured confounding. We illustrate the proximal framework via simulation studies and a variety of data applications in the health and social sciences.