Nonparametric Covariates Adjustment for Youden Index

• Activity Number: 407
• Type: Contributed
• Date/Time: Tuesday, August 2, 2011 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Haochuan Zhou*+
• Companies: Georgia State University
• Keywords: Youden Index ; heteroscedastic regression model ; local polynomial regression ; working sample ; asymptotic normality ; asymptotic consistency

Abstract

The receiver characteristic curve (ROC) has been a popular method to evaluate the accuracy of a diagnostic test. Summaries of ROC, the area under the ROC curve and the Youden Index (YI) are two main indexes to measure the capability of a test. Sometimes, the known information from some covariates may influence the accuracy of a test. We propose nonparametric method for covariate adjustment of the YI. Heteroscedastic models under normal and non-normal error assumption are developed and investigated. The local polynomial regression is applied to estimate mean and variance functions of the model. In the model which without normality assumption for the error term, we propose an empirical estimator for YI which efficaciously utilizes available data to construct working samples at any covariate value of interest and is computationally efficient. Under two assumptions, we explore the asymptotic normality, strong uniform convergence rate and asymptotic consistency for the YI estimators. Plenty of simulations are conducted to illustrate the effectiveness for both estimate and the robustness for the empirical estimate. A real data of diabetes disease study is used to demonstrate the methods.