A1146
Title: Objective Bayes model selection for the Behrens-Fisher problem
Authors: Marco Ferreira - Virginia Tech (United States) [presenting]
Abstract: The aim is to propose a novel objective Bayes model selection approach for the Behrens-Fisher problem. The Behrens-Fisher problem concerns testing the equality of means of two populations when their variances are different. Typically, the data analyst would first test for the equality of the variances and, if the null hypothesis of such equality was rejected, the analyst would apply a test of equality of the means assuming unequal variances. Instead of this two-step approach, an omnibus objective Bayes solution that at the same time tests equality of means and of variances is developed. Simulation studies provide a comparison of the performance of our solution versus the performance of previously proposed methods. Applications to publicly available datasets illustrate the usefulness of the proposed approach.
