A Comparison of Empirical Bayes and Reference Prior Methods for Spatio-Temporal Data Analysis

In Bayesian analysis of spatio-temporal data, the problem of selecting prior distribution for model parameters is of great demand. This paper considers two most popular approaches, empirical Bayes and reference prior, for Bayesian inference. We then use simulation to compare the frequentist properties of these two methods. Since, posterior propriety of the reference prior is only established under separable correlation models, this comparison is concentrated on this case.