On estimating the area under the ROC curve in ranked set sampling

Area of Science:

• Biostatistics
• Medical Informatics
• Diagnostic Accuracy

Background:

• Receiver operating characteristic (ROC) curves are standard for evaluating continuous biomarkers in medical research.
• The area under the ROC curve (AUC) quantifies overall biomarker performance.
• Ranked set sampling (RSS) offers potential efficiency gains over simple random sampling.

Purpose of the Study:

• To develop and evaluate novel estimators for the area under the ROC curve (AUC) specifically designed for ranked set sampling (RSS).
• To compare the performance of the proposed AUC estimators against existing methods.

Main Methods:

• Development of three new AUC estimators for RSS data.
• The first estimator assumes data normality.
• The other two estimators utilize a Box-Cox transformation followed by parametric or kernel-density-based approaches.

Main Results:

• A simulation study demonstrated that the new RSS-based AUC estimators provide advantages in specific scenarios.
• The proposed methods showed competitive or superior performance compared to existing literature estimators under certain conditions.
• The utility of the new estimators was validated using real medical data.

Conclusions:

• The developed AUC estimators offer valuable alternatives for biomarker performance evaluation in RSS settings.
• These methods enhance the accuracy and efficiency of biomarker assessment in medical research.
• The findings support the application of these novel estimators in clinical practice and further research.