COMPARING SEVERAL DIAGNOSTIC PROCEDURES USING THE INTRINSIC MEASURES OF ROC CURVE

Comparison of diagnostic tests is essential in medicine. Test procedures for comparing two or more ROC curves are all based on measures d', AUC and the maximum likelihood estimates of binormal ROC curves. However, intrinsic measures such as sensitivity and specificity also play a pivotal role in assessing the performance of several diagnostic procedures. In this paper, a new methodology is proposed in order to compare several diagnostic procedures using the intrinsic measures of ROC curv