Performance Comparison of Independence Tests in Two-Way Contingency Table

Abstract

Several test statistics are available for testing the independence of categorical variables from two-way contingency tables. A vast majority of published articles used the Pearson’s chi-squared test for such purposes; however, this test statistic may lead to biased conclusions under certain conditions. Therefore, we aimed to compare the performance of test statistics via a comprehensive simulation study considering several factors in contingency tables. We also evaluated the performance of each test statistic on a real-life dataset. This study contributes to the literature guiding researchers to select an appropriate test statistic under different conditions.

Keywords

• contingency table
• power
• power divergence
• simulation study
• type one error