A Study of Models for High-Dimensional Spatial Binary Data

• Activity Number: 91
• Type: Invited
• Date/Time: Sunday, August 9, 2015 : 9:30 PM to 10:15 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Yawen Guan* and Murali Haran
• Companies: Penn State and Penn State
• Keywords: SGLMMs ; GMRF ; GP ; Binary

Abstract

Modeling high-dimensional spatial binary data using traditional spatial generalized linear mixed model (SGLMM) is computationally intensive. Moreover it inflates the variance of fixed effect (regression coefficient) estimates. In this study we explore methods that reduce the computational cost while also alleviating the confounding issue between fixed and random effects. We study, via simulated examples, approaches based on representing the high-dimensional spatial random effects by reduced-dimensional random vectors. Taking into account flexibility and computational cost, we work toward an approach that works well for both discrete-domain (Gaussian Markov random field) models as well as continuous-domain (Gaussian process) spatial models.