| Session Title: Evidence Synthesis using Bayesian Mixture Models |
| Demonstration Session 795 to be held in Panzacola Section H2 on Saturday, Nov 14, 11:50 AM to 12:35 PM |
| Sponsored by the Quantitative Methods: Theory and Design TIG |
| Presenter(s): |
| J Michael Menke, University of Arizona, menke@u.arizona.edu |
| Abstract: Bayesian Mixture Modeling (BMM) offers a way to compare treatments or intervention when head-to-head trials have never been done and unlikely or too expensive to ever be done. Research and expertise domains tend to stay within a single field or discipline. As such, a domain may build more and deeper knowledge of the same domain, but applied and policy questions, which require comparison between systems. Treatment decisions may be loosely judged through meta-analytic techniques that estimate effect sizes along with study quality. The more practical issue of estimating the extent of treatment advantage can yield additional information that can be used in cost-effectiveness and other factors that help reduce uncertainty and improve decisions. BMM is a method of data synthesis that allows studies from different specialties and programs to be compared as if they are all arms of a single study, with certain convenient advantages. |