Abstract
The thesis is about the maximum likelihood estimation for spatial random effects model. The maximum likelihood estimation has no closed form but the numerical solution can be effectively solved through the EM algorithm. Moreover, some regularization methods for covariance parameters are incorporated in the estimation procedure to further simplify the fitted model structure. The simulation results verify that the proposed estimation methods have good performance in both estimation and prediction under various non-stationary models. The methodology is also applied to a real data set, chlorophyll concentration data from SeaWiFS projects, for illustration.Keywords : fixed rank kriging, MLE, EM algorithm, graphical lasso