Tapering spatio temporal models

We consider regression models for multivariate spatio temporal data. We view the data as a time series of spatial processes and work in the setting of dynamic models. In order to add exibility we consider regression models with spatio temporal varying coefficients. Spatial dependence among the di fferent measurements is attained considering the linear model of coregionalization. Since spatio-temporal data are tipically of large dimension we propose to perform estimation both through maximum likelihood by means of the EM algorithm and a modi ed version of it exploiting the covariance tapering likelihood function.