A Stabilized and Versatile Spatial Prediction Method for Geostatistical Models

Abstract

A Stabilized and Versatile Spatial Prediction Method for Geostatistical Models Chun-Shu Chen1 1Institute of Statistics and Information Science, National Changhua University of Education, Taiwan ROC ? ?: Geostatistical models are often used to predict spatial variables of interest, but the parameters of the spatial correlation structure usually cannot be well estimated, resulting in the spatial predictor that may be unstable. In this talk, we apply a data perturbation procedure to stabilize the spatial predictor, which is not only continuous but also differentiable with respect to the response variables even after plugging-in the estimated model parameters. Moreover, it is known that different spatial predictors obtained from different methods generally have different performances under different situations. To avoid confronting a model selection process, in this talk, we will also propose a weighted estimation method based on Stein’s unbiased risk estimate for averaging over a sequence of candidate spatial predictors, leading to a versatile spatial predictor that is adaptive to the underlying spatial process. Some numerical experiments are performed for illustrating the superiority of the proposed method and an application of a real data set is also presented. ?