Weighted Wilcoxon estimate for the covariate-specific ROC curve

 
   Lead Investigator:    Xiaogang Duan
   Institution      :    Beijing Normal University, Beijing, China
   E-Mail           :    xgduan@bnu.edu.cn
   Proposal ID      :    445



   Proposal Description:
   The covariate-specific receiver operating characteristic (ROC) curve is an important tool for evaluating the classification accuracy of a diagnostic test when it is associated with certain covariates. In this project, a weighted Wilcoxon estimate is constructed for estimating the ROC curve under the framework of location scale model for the test result. Asymptotic normality is established, both for estimators for the regression parameter and the covariate-specific receiver operating characteristic curve at a fixed false positive point. Simulation results show that the Wilcoxon estimator compares favorably to its main competitors in terms of the standard error, especially when the distribution is heavy tailed or outliers exist both in the response and the covariates. We also apply the new procedure to the dementia data from the national Alzheimer's coordinating  center.