The Evaluation of Integrals of the Form [f(s,t)*Bivariate Normal(s,t)]. Application to Correlated Bivariate Logistic-Gaussian Models

• Activity Number: 242 - Contributed Poster Presentations: Biometrics
• Type: Contributed
• Date/Time: Monday, July 31, 2017 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Edmund Ameyaw* and Paul Bezandry and Victor Apprey and John Kwagyan
• Companies: Howard University and Howard University and Howard University and Howard University College of Medicine, General Clinical Research Center
• Keywords: Correlated Data ; Logistic -Gaussian Distribution ; Maximum Marginal Likelihood ; Bivariate Response ; Gauss-Hermite quadrature

Abstract

The Logistic-Gaussian distribution is used in statistical applications to account for clustering among binary outcomes. However, its extension and applicability to bivariate outcomes is limited. We developed a model for correlated bivariate binary data that incorporated the Logistic-Gaussian distribution. A bivariate normally distributed variate is decomposed into a product of two univariate normally distributed variate and applied to the development of a correlated bivariate logistic Gaussian model. Bivariate response probabilities in terms of random effects models are formulated, and maximum marginal likelihood estimation procedures based on Gauss-Hermite quadrature are used. Application to the analysis of vision loss in diabetic retinopathy is discussed.