Generalized Sparse Precision Matrix Selection for Fitting Multivariate Gaussian Random Fields to Large Data Sets

摘要

We present a new method for estimating multivariate, second-order stationaryGaussian Random Field (GRF) models based on the Sparse Precision matrixSelection (SPS) algorithm, proposed by Davanloo et al. (2015) for estimatingscalar GRF models. Theoretical convergence rates for the estimatedbetween-response covariance matrix and for the estimated parameters of theunderlying spatial correlation function are established. Numerical tests usingsimulated and real datasets validate our theoretical findings. Datasegmentation is used to handle large data sets.