Within-Cluster Resampling Methods for Clustered ROC Data

• Activity Number: 561
• Type: Contributed
• Date/Time: Wednesday, August 6, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Zhuang Miao*+ and Liansheng Larry Tang
• Companies: and George Mason University
• Keywords: ROC Curve ; Clustered ROC Data ; within-cluster Resampling

Abstract

The diagnostic studies in which each patient has several normal and abnormal observations generate clustered ROC data. Within the same cluster, observations are naturally correlated, and the cluster size may be random. The traditional ROC methods on clustered data can result in a biased variance estimator and subsequently lead to incorrect statistical inference. We introduce resampling methods on clustered ROC data to account for the within-cluster correlation. The within-cluster resampling ROC methods work as follows. First, one observation is randomly selected from each patient/cluster, and then the traditional ROC methods are applied on the resampled data to obtain resampled ROC estimates. These steps are performed many times and the average of resampled ROC estimates is the final estimator. The proposed method does not require a specific within-cluster correlation structure and yields a valid estimator while accounting for the within-cluster correlation. We compare the proposed method with existing methods in extensive simulation studies and apply the proposed method to two eye rating examples.