Exact Inference for the Common Odds Ratio in a Series of 2-by-2 Tables and Its Applications to the Cases of Rare Events

• Activity Number: 139
• Type: Contributed
• Date/Time: Monday, August 1, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Dungang Liu*+ and Regina Y. Liu and Minge Xie
• Companies: Rutgers University and Rutgers University and Rutgers University
• Keywords: meta-analysis ; continuity correction ; rare events ; exact inference ; odds ratio ; zero total study

Abstract

This paper proposes a simple meta-analysis method for the common odds ratio in a series of independent 2 by 2 tables based on combining confidence distribution (CD) framework. Almost all commonly used meta-analysis methods rely on large sample approximation. When dealing with rare events, such approximation is nevertheless far from appropriate. In addition, these methods often requires artificial imputing to zero cells, which is known to have an uncertain impact on final conclusions. To circumvent these pitfalls, we propose to combine the exact test result of each study. The combined result is summarized as a function over the parameter space, which is easy to use for overall exact inference. This novel approach enables us to incorporate the studies having zero cells without artificial imputing. Asymptotic efficiency of the proposed approach is established. Numerical studies using simulated and real data show its superiority over existing popular methods.