An Exact Density-Based Empirical Likelihood Ratio Test for Paired Data.

Area of Science:

• Statistics
• Biostatistics
• Nonparametric Methods

Background:

• The Wilcoxon rank-sum test is a powerful nonparametric method for comparing two groups with paired data.
• Existing methods can be suboptimal or fail with skewed distributions or non-constant shifts.

Purpose of the Study:

• To introduce a novel empirical likelihood (EL) ratio approach for testing the equality of marginal distributions in bivariate populations.
• To demonstrate the superiority of the proposed EL test over traditional nonparametric procedures.

Main Methods:

• Development of an exact empirical likelihood (EL) ratio test for comparing marginal distributions.
• Extensive Monte Carlo simulations to evaluate the test's performance under various shift alternatives.
• Application of the EL ratio test to real-world medical study data.

Main Results:

• The proposed EL ratio test demonstrates superior performance compared to classic nonparametric tests in shift alternative scenarios.
• The EL test is particularly effective when data distributions are skewed or exhibit non-constant shifts.
• Monte Carlo studies confirm excellent operating characteristics for the proposed method.

Conclusions:

• The empirical likelihood (EL) ratio test provides a robust and powerful alternative for comparing two groups from continuous bivariate populations.
• This method is especially advantageous in situations where traditional nonparametric tests may be inadequate.
• The test's efficacy is validated through simulations and real-world medical data analysis.