Journal of Statistics: Advances in Theory and ApplicationsAuthor's: Chrysoula Dimitriou-Fakalou
Pages: [11] - [44]
Received Date: August 16, 2009
Submitted by:
For observations recorded regularly on a perfect circle or sphere, we present a new class of stationary models that clothe naturally the spatial second-order dependence. At first, the new model might be seen as the cyclical analogue of the standard auto-regressive model. Further and due to the spherical structure, a finite auto-regressive as well as finite moving-average dependence may be clothed simultaneously via the so-called CAL model. Some existing results from time series might be simplified and we may easily proceed with the estimation and statistical inference for the coefficients of best linear prediction, which are needed for kriging on the spherical surface.
auto-correlation, auto-linear model, best linear prediction coefficient.
