Model Selection for a Class of Spatio-temporal Models for Areal Data

Abstract. We present a method to perform model selection based on predictive density in a class of spatio-temporal dynamic generalized linear models for areal data. These models assume a latent random field process that evolves through time with random field convolutions; the convolving fields follow proper Gaussian Markov random field processes. Parameter and latent process estimation based on Markov Chain Monte Carlo and the forward information filter backward sampler, respectively, is showed. Finally, an application using several specifications of the general model on homicide data in the State of Esp´ırito Santo is presented showing the results of model selection.

• Type: Conference
• Year: 2007
• Where: GEOINFO
• Authors: Juan C. Vivar, Marco A. R. Ferreira
• Keywords: GEOINFO 2007, Information Technology, Markov Chain Monte Carlo, Model Selection, Random field Process