Semiparametric Bayesian Model for Areal Data with Space-Time Varying Coefficients

• Activity Number: 184
• Type: Contributed
• Date/Time: Monday, August 1, 2011 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Bo Cai*+ and Andrew B. Lawson and Monir Hossain and Jungsoon Choi
• Companies: University of South Carolina and Medical University of South Carolina and The University of Texas and Medical University of South Carolina
• Keywords: Bayesian regression ; hierarchical structure model ; Dirichlet process ; Space-time models

Abstract

In spatial analysis, the effects of covariates on the outcome are usually assumed to be invariant across areas. However, the spatial configuration of the areas may potentially depend on not only the structured random intercept but also spatially varying coefficients of covariates. In addition, the distribution of spatially varying coefficients are not always normally distributed. In this case, the normality assumption could lead to potential biases of estimations. In this article, we propose a semiparametric space-time model from a Bayesian perspective. The spatially varying coefficients of space-time covariates are modeled by using the spatial Dirichlet process prior which yields data-driven deviations from the normality assumption. The proposed semiparametric approach evinces the improvement of prediction compared to usual Bayesian spatial-temporal models with normality assumption on spatial-temporal random effects and to the model with the spatial random intercept modeled nonparametrically. A simulation example is presented to evaluate the performance of the proposed approach with the competing models. A real data example is used for an illustration.