Two-Sample Empirical Likelihood-Based Tests for Mean: From Frequentists to Bayesian Type Techniques with Applications to Case-Control Studies

• Activity Number: 229
• Type: Contributed
• Date/Time: Monday, August 4, 2014 : 2:00 PM to 3:50 PM
• Sponsor: International Indian Statistical Association
• Author(s): Ge Tao*+ and Albert Vexler
• Companies: University at Buffalo and University at Buffalo
• Keywords: Empirical likelihood ; Bayes factor ; Maximum likelihood estimation ; Two-sample tests

Abstract

Many clinical experiments are designed to be in a form of case-control studies for detecting discriminating ability of biomarkers or comparing treatment effects. Two-sample statistical tests are common procedures applied in such investigations. Avoiding parametric assumptions regarding data distributions, we consider different forms of empirical likelihood ratio tests to be employed in case-control studies. In this paper, we also proposed and evaluate novel two-sample empirical likelihood ratio based tests that involve Bayes factor type mechanisms in a nonparametric manner. The asymptotic properties of the proposed techniques are presented. We employ an extensive Monte Carlo study to evaluate the theoretical results as well as to compare the power of various two-sample tests for mean. The applicability of the proposed methods is demonstrated via a study associated with myocardial infarction disease.