Nonparametric Estimation of the Covariate-Specific ROC Curve]{Nonparametric Estimation of the Covariate-Specific ROC Curve in Presence of Ignorable Verification Bias

• Activity Number: 618
• Type: Topic Contributed
• Date/Time: Thursday, August 5, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Health Policy Statistics Section
• Author(s): Danping Liu* and Xiao-Hua Zhou+
• Companies: University of Washington and University of Washington
• Keywords: Alzheimer's disease ; covariate-specific ROC curve ; ignorable missingness ; verification bias ; weighted estimating equations

Abstract

Covariate-specific ROC curves are often used to evaluate the classification accuracy a medical diagnostic rest, when the accuracy of the test is associated with the covariates. In cohort studies, the gold standard is subject to missingness due to high cost or harmfulness to the patient. We propose a nonparametric estimation of the covariate-specific ROC curves with a partial missing gold standard. The framework is built on a location-scale model for the test result, with the unspecified residual distributions. Under the missing at random (MAR) assumption, three ROC curve estimators are proposed and compared, namely, imputation-based, inverse probability weighted and doubly robust estimators. The asymptotic normality of the estimated ROC curve is proved, and the point-wise confidence interval is also derived. The proposed method is applied to the data in Alzheimer's disease research.