A Semi-Local Likelihood Regression Method for Group Testing Data

• Activity Number: 63
• Type: Topic Contributed
• Date/Time: Sunday, August 4, 2013 : 4:00 PM to 5:50 PM
• Sponsor: Biometrics Section
• Author(s): Dewei Wang and Haiming Zhou*+ and Karunarathna B. Kulasekera
• Companies: Clemson University and University of South Carolina and University of Louisville
• Keywords: Nonparametric binary regression ; Local smoothing ; Pooled data

Abstract

Group (pooled) testing has proven to be an efficient method of reducing the cost and time associated with screening individuals for infectious diseases.It proceeds to test pooled specimens formed from amalgamating specimens collected from individuals, rather than testing them one-by-one. Many authors have proposed methods that make use of the observed pool testing results for the purposes of estimating population level characteristics. Recently, regression methods that relate individual level covariates to the observed pool responses have been proposed. Most of these methods rely on parametric modeling assumptions. In fact, to our knowledge, only two nonparametric regression methods for pool response data exist, but neither of the proposed estimators enjoys asymptotic normality. In this talk, I will introduce a new nonparametric estimation method based on a semi-local weighted log-likelihood, and discuss the asymptotic properties of our proposed estimator. We illustrate our nonparametric regression methodology using hepatitis B data obtained from the National Health and Nutrition Examination Survey.