Model Selection in Spatial Regressions with Power Exponential Covariance Family

Keywords: Cross Validation, Kriging, Power Exponential Covariance

Abstract: A strategy to select the best model in a family of the spatial regression model with power exponential covariance is presented. The model being considered has been used for spatial data as well as for computer experiments. This stepwise strategy deals with the linear model terms and the covariance terms simultaneously to find the best combination of nonzero coefficients in both terms. It is compared to the criterion of cross validation, AIC and BIC by numerical experiments.