Incorporating Retests Into Group Testing Regression Model Estimation

• Activity Number: 31
• Type: Contributed
• Date/Time: Sunday, July 29, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Boan Zhang*+ and Christopher Bilder and Joshua Tebbs
• Companies: University of Nebraska and University of Nebraska and University of South Carolina
• Keywords: Binary response ; Generalized linear model ; EM algorithm ; Group testing ; Prevalence estimation ; Relative efficiency

Abstract

Group testing has become a standard procedure to reduce the costs of screening a large number of individuals for rare diseases. In many applications of group testing, the goal often includes both estimating the prevalence of the disease and identifying individuals as positive or negative. In addition to the initial group tests, more individual or group tests (called "retests") are performed in these studies to decode positive groups. In this paper, we investigate how regression models can be fit to group testing data from three commonly used group testing protocols. Simulation evidence is presented to show significant efficiency gains of incorporating retests into the estimation process, as compared to using the initial group test results alone. In addition, the number of tests per unit information is used as a criterion to determine the best group testing protocol.