Group Testing Regression Models for Multiple Traits

• Activity Number: 359
• Type: Contributed
• Date/Time: Tuesday, August 2, 2011 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Boan Zhang*+ and Christopher Bilder and Joshua M. Tebbs
• Companies: University of Nebraska at Lincoln and University of Nebraska at Lincoln and University of South Carolina
• Keywords: correlated binary data ; expectation-solution algorithm ; generalized estimating equations ; latent response ; pooled testing ; unobserved response

Abstract

Group testing, where groups of individual specimens are composited to test for a binary trait (e.g., infectious disease), is a procedure commonly used to reduce the costs of screening a large number of individuals. Group testing data is unique in that only group responses may be observed, but inferences are necessary at the individual level. A further challenge arises when specimens are screened for multiple traits leading to unobserved correlated binary responses for the individuals. In our poster, we propose the first regression models for these types of responses. Through the use of generalized estimating equations and the expectation-solution algorithm, we develop methodology to fit models when only the group responses are available. An important consequence of our methodology is that the proposed regression models are easily adapted to a longitudinal testing situation where individual subjects appear in the same groups over time. Simulation studies are performed to evaluate and compare small sample performance of our methods. Finally, the proposed modeling procedures are applied to chlamydia and gonorrhea screening data collected as part of the Infertility Prevention Project.