Comparison of causal and non-causal linear prediction models for multidimensional spectrum estimation

The authors classify and compare many of the linear prediction models that have appeared recently in the literature. The models considered are classified as quarter-plane causal, half-plane causal, three-quadrant causal, and noncausal. Comparisons are made of the corresponding notions of innovations, regularity, and singularity, and the possibility of Wold-like decompositions for each is discussed. The present work is motivated by the use of these linear prediction schemes as models for rational, multidimensional spectrum estimation