Selection of rank and basis functions in the Spatial Random Effects Model
Use of the Spatial Random Effects (SRE) model has been motivated by the need to reduce the amount of computation involved in obtaining optimal predictors and their prediction standard errors. This reduction in computation relies on the covariance matrix of the random effects being of low rank. In this proceedings paper, the selection of the rank and the selection of the accompanying basis functions of the SRE model are investigated. We explore selection criteria that include the generalized degrees of freedom and two information criteria; their relative performance is assessed by simulation.