A Permutation Test for Comparing Paired Roc Curves

Receiver Operating Characteristic (ROC) analysis has been used as a popular technique of comparing the performance of two paired diagnostic tests data while the Area Under the Curve denoted as AUC which also represents the probability that test result due to diseased subject is greater than test result due to nondiseased subject can be used to summarize the overall activities between two ROC curves. This paper was aimed at assessing a difference in the AUCs of a continuous matched pair of data as well as tackling the problem of exchangeability of the labels between two diagnostic tests within subjects which characterizes previous studies. The work proposed a Modified Wilcoxon Signed-Rank Test (MWSRT) by permuting between subject labels within each diagnostic test and adjusting for the possible presence of tied absolute values of differences as well as equating the resulting test statistic to difference in AUCs.