Nonparametric Bayesian Methods for Measurement Error Problem in Matched Case-Control Studies

• Activity Number: 303
• Type: Contributed
• Date/Time: Tuesday, August 3, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Nels Johnson*+ and Inyoung Kim
• Companies: Virginia Tech and Virginia Tech
• Keywords: Conditional likelihood ; Dirichlet process ; Matched case-control ; Nonparametric Bayesian

Abstract

In epidemiological research, matched case-control studies are popular. Measurement error of covariates is a modeling problem that often needs to be addressed. However, there is no statistical approach to handle both problems together. We propose a nonparametric Bayesian method for measurment error models in matched case-control studies, modeling the measurment error as a Dirichlet process mixture. We compare our method with a parametric Bayesian method and a frequentist conditional likelihood method. We show that our method performs better using a simulation study. We also demonstrate the advantages of our approach on a 1-5 matched case-crossover study in public health epidemiology.