PREDICTION DIAGNOSTICS FOR SPATIAL LINEAR-MODELS

Case deletion diagnostics are developed for detecting observations that are influential for prediction in linear models with a general covariance matrix. A primary application of such results is in universal kriging and the related methodologies of ordinary kriging and intrinsic random function kriging. In these applications the linear model and the covariance matrix of the data are determined by the locations at which observations are taken. Computational formulae are given that make the procedures feasible. The diagnostics are illustrated using an example.