主题：美国Georgia州立大学秦更生博士统计学术报告

时间：2008年5月14日下午4:30
地点：数理学院309会议室

报告题目: Bootstrap and Empirical Likelihood-based Nonparametric Inference For the Difference between Two Partial AUCs

摘要: Comparing the accuracy of two continuous-scale diagnostic tests is increasingly important when a new test is developed.
Traditional way of comparing entire areas under two Receiver Operating Characteristic (ROC) curves is not sensitive when two ROC curves cross each other. Comparing the areas under two ROC curves over a specific interval of false positive rates is a more appropriate way to evaluate the accuracy of two diagnostic tests. In this paper, we have derived a normal approximation for the distribution of the difference between two estimated partial areas under the ROC curves (partial AUCs). The empirical likelihood ratio for the difference between two partial AUCs is defined and it is shown that its asymptotic distribution is a scaled chi-square distribution. Bootstrap and empirical likelihood based inferential methods are proposed for the difference between two partial AUCs.  Five confidence intervals for the difference between two partial AUCs are constructed based on the normal and chi-square approximations.
Simulation studies are conducted to compare the finite sample performance of the proposed intervals. A real example is used to illustrate the application of the recommended intervals.