Jeffrey A. Boatman, a Postdoctoral Associate in the Division of Biostatistics at the University of Minnesota, will present:

“One Out of Many: Combining Data Sources for Causal Inference”

Abstract: The increasing multiplicity of data sources offers exciting possibilities for causal inference, particularly if observational and experimental sources can be combined for estimation. Borrowing between data sources can potentially result in more efficient estimators, but it must be done in a principled manner to minimize bias. Furthermore, when the effect of treatment is confounded, as in observational data sources or in clinical trials with noncompliance, special care must be taken to appropriately adjust for confounding. In this talk I consider the problem of estimating causal effects from a primary source and borrowing from any number of supplemental sources. I propose using regression-based estimators that borrow by sharing regression coefficients and parameters between data sources. Multisource exchangeability models govern the extent of borrowing through Bayesian model averaging. Simulation results show that a Bayesian linear model and Bayesian additive regression trees both have desirable properties and borrow under appropriate circumstances. I give special attention in this talk to the regulatory tobacco research that motivated development of the method.

A social tea will be held at 3:00 p.m. in A434 Mayo. All are Welcome.