Joint Estimation of Multiple Disease-Specific Sensitivities and Specificities via Crossed Random Effects Models for Correlated Reader-Based Diagnostic Data

• Activity Number: 562
• Type: Contributed
• Date/Time: Wednesday, August 6, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Alex de Leon*+ and Niroshan Withanage and Christopher Rudnisky
• Companies: University of Calgary and Sabaragamuwa University of Sri Lanka and University of Alberta
• Keywords: Binocular data ; Correlated binary outcomes ; Latent variable ; Likelihood estimation ; Multivariate probit model ; Tetrachoric correlations

Abstract

We present a model for describing correlated binocular data in a reader-based diagnostic study where the same group of readers evaluates the presence or absence of certain pathologies in the paired organs of patients. Crossed random effects are incorporated to meaningfully delineate various associations in the data. To overcome the computational complexity in the evaluation and maximization of the marginal likelihood, we adopt the data cloning approach, which provides maximum likelihood estimates under the Bayesian paradigm. The efficiency of the estimates is assessed in a simulation study. We apply our model to data from a diabetic retinopathy study.