Dynamic Spatio-Temporal 'Auto' Models with Third-Order Dependence

• Activity Number: 264
• Type: Contributed
• Date/Time: Monday, August 10, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Xiaomao Xia* and Athanasios Micheas and Christopher K. Wikle
• Companies: University of Missouri and University of Missouri and University of Missouri
• Keywords: MRF ; Spatio-temporal model ; Higher-order dependence ; Hierarchical model

Abstract

Spatio-temporal models are ubiquitous in scientific applications. In many of these situations, conditional Gaussian models with linear (two-way) interaction may not be appropriate. An alternative is to consider the class of "auto" models (e.g., auto-logistic, auto-Poisson, auto-Gaussian) in which case the dependence is built directly in terms of Markov random field theory through conditional interactions. Outside of the Gaussian case, where quadratic nonlinear dynamic spatio-temporal models have been considered, very little work has been done to investigate interaction orders higher than two in spatio-temporal dynamic auto models in the broader exponential family. Here we consider an extension of the class of spatio-temporal auto models to include third-order interactions. We demonstrate our results on simulated data and real-world environmental applications.