Epidemiologists often study the effect of an exposure on an outcome to identify how manipulating the exposure in some population will affect incidence of the outcome. The emerging field of causal inference attempts to formalize the assumptions required to identify these causal effects and to develop new methods to estimate effects when standard methods fail. As in other fields, Bayesian methods can be used in causal inference to incorporate prior knowledge or to improve the performance of estimation procedures. Nonetheless, the majority of formal causal-inference methods in health and medical science have been frequentist. This symposium will highlight the work of 3 speakers who have employed Bayesian methods to improve causal inference in epidemiology. The discussant will tie together the major themes of the talks and connect them to existing issues in epidemiology.
Session Chair: Jess Edwards, University of North Carolina at Chapel Hill
Infinity plus two: robustness the Bayesian way Francesca Dominici, Harvard University
Can There be Bayesian Double Robustness? Olli Saarela, University of Toronto
