Confidence Intervals for AUC and pAUC by Empirical Likelihood.

Area of Science:

• Biostatistics
• Medical Diagnostic Test Evaluation

Background:

• Area Under the Receiver Operating Characteristic Curve (AUC) and Partial AUC (pAUC) are key metrics for medical diagnostic test performance.
• Existing methods for AUC and pAUC estimation often involve complex variance calculations.

Purpose of the Study:

• To propose a novel two-sample empirical likelihood approach for hypothesis testing and confidence interval construction for AUC and pAUC.
• To offer a statistically robust and computationally simpler alternative to existing methods.

Main Methods:

• A two-sample empirical likelihood ratio test is developed for nonparametric settings.
• The test statistic asymptotically follows a chi-square distribution under the null hypothesis.
• This method avoids the need to estimate complex scale factors or variances.

Main Results:

• Simulations demonstrate the proposed method's superior performance compared to competitors across various scenarios.
• The empirical likelihood ratio test provides an accurate chi-square distribution, simplifying statistical inference.
• Real-world data examples with accompanying R code are provided to illustrate practical application.

Conclusions:

• The proposed empirical likelihood method offers a competitive and efficient approach for statistical inference of AUC and pAUC.
• This method simplifies the process of evaluating medical diagnostic tests, enhancing their reliable application.
• The study provides valuable tools for researchers and practitioners in medical diagnostics.