Estimating AUC Using Test-Dependent Sampling

• Activity Number: 658
• Type: Contributed
• Date/Time: Thursday, August 2, 2012 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Bethany Horton*+ and Haibo Zhou
• Companies: The University of North Carolina at Chapel Hill and The University of North Carolina at Chapel Hill
• Keywords: Test dependent sampling ; Area under ROC curve (AUC) ; Biased sampling ; Missing data

Abstract

Area under the receiver operating curve (AUC) is a summary measure used to describe the ability of a screening test to discriminate between diseased and non-diseased subjects. As evaluating the true disease status can be costly, it is beneficial for researchers to increase study efficiency by allowing selection probabilities to depend on the screening test. We consider a test dependent sampling (TDS) design where TDS inclusion is dependent on a continuous screening test measure. To improve efficiency, this sampling design incorporates a simple random sample and the supplemental TDS component, as opposed to a design using only a simple random sample of the same size. Disease status is known only for those included in the two sampling components. We propose an AUC estimator which incorporates the TDS, a simple random sample, and the remaining un-sampled portion of the population whose disease status is unknown. This estimator gives improved estimates of the AUC compared to an estimate using only a simple random sample of the same size.