Fast Maximum Likelihood Inference for Spatial Generalized Linear Mixed Models

• Activity Number: 83 - Your Invited Poster Evening Entertainment: No Longer Board
• Type: Invited
• Date/Time: Sunday, July 30, 2017 : 8:30 PM to 10:30 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Yawen Guan* and Murali Haran
• Companies: Penn State University and Pennsylvania State University

Abstract

Spatial non-Gaussian data are common in many environmental disciplines. Spatial generalized linear mixed models (SGLMMs) are flexible models for such data but they are computationally intensive. Moreover, SGLMMs inflate the variance of fixed effect (regression coefficient) estimates. We explore fast maximum likelihood inference for methods that reduce the computational cost while also alleviating the confounding issue between fixed and random effects. We study a projection-based methodology that represents the high-dimensional spatial random effects by reduced-dimensional random vectors. We develop a Monte Carlo Expectation Maximization (MC-EM) algorithm for efficient inference for these models. Our approach applies to both discrete-domain (Gaussian Markov random field) as well as continuous-domain (Gaussian process) spatial models.