Confidence Intervals for the Difference Between Two Partial AUCs

• Activity Number: 65
• Type: Topic Contributed
• Date/Time: Sunday, July 29, 2012 : 4:00 PM to 5:50 PM
• Sponsor: Biometrics Section
• Author(s): Xin Huang*+ and Gengsheng Qin and Yan Yuan and Xiao-Hua (Andrew) Zhou
• Companies: Fred Hutchinson Cancer Research Center and Georgia State University and Georgia State University and University of Washington
• Keywords: Bootstrap ; Confidence interval ; Empirical likelihood ; ROC curve

Abstract

With new diagnostic tests being developed and marketed, the accuracy comparison of diagnostic tests is of high importance. An effective method to evaluate the difference between the diagnostic accuracy of two tests is to compare partial areas under the receiver operating characteristic curves (AUCs). In this paper, we review parametric methods and study semi-parametric and non-parametric methods for the difference between two partial AUCs. Under each model assumption, for the difference between two partial AUCs, we derive a normal approximation for its distribution, define an empirical log-likelihood ratio and show that it follows a scaled chi-square distribution. We construct five confidence intervals for the difference based on normal approximation, bootstrap, and empirical likelihood methods. In addition, we perform extensive simulation studies to compare the finite sample performances of these intervals, and use a real example as an application of our recommended intervals. The analysis results indicate that the proposed hybrid bootstrap and empirical likelihood intervals outperform other intervals in most cases.