B1065
Title: Bayesian semiparametric approaches to tailoring strategies
Authors: David Stephens - McGill University (Canada) [presenting]
Abstract: Frequentist semiparametric methods for optimizing and tailoring treatments to patient characteristics are well established, but typically rely on asymptotic justifications, for example, sandwich estimation, for uncertainty representations. We will present computational Bayesian methods that give exact posterior credible intervals (under an assumed flexible semiparametric specification) in the treatment-tailoring problem. These Monte Carlo-based approaches rely on a Bayesian nonparametric formulation, and can be applied in regression-based and inverse weighting approaches.
