Gaussian Copula Processes in Spatial Generalized Linear Models

• Activity Number: 409 - Bayesian Space-Time Modeling
• Type: Contributed
• Date/Time: Tuesday, July 30, 2019 : 2:00 PM to 3:50 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Robert Richardson*
• Companies: Brigham Young University
• Keywords: Stochastic Process; Spatial Models; Non-Gaussian Models; Spatial Mortality

Abstract

A spatial generalized linear model is presented where the spatial correlation comes from a Gaussian copula process instead of from spatially varying parameters in the link function. This method allows for straight-forward spatial GLMs in cases where a link function is not easily established, such as a personalized discrete distribution or distributions that are modified at 0. The results are compared theoretically and empirically to existing method and an application in spatial mortality rates is presented.