J Austral Math Soc Ser B 33 pp192--210, 1991.

Spectral factorisation and prediction of multivariate processes with time-dependent rational spectral density matrices

N. M. Spencer and V. V. Anh

Abstract

This paper considers discrete multivariate processes with time-dependent rational spectral density matrices and gives a solution to the spectral factorisation problem. As a result, the corresponding state space representation for the process is obtained. The relationship between multivariate processes with time-dependent rational spectral density matrix functions and multivariate ARMA processes with time-dependent coefficients is discussed. Solutions for the prediction problem are given for the case when only finite data is available and the case when the whole history of the process is known.

Browse the article

Authors

N. M. Spencer

V. V. Anh

School of Mathematics, Queensland Institute of Technology, Brisbane Q 4001, Australia.