Modeling of Multiple Time Series by the Method of Successive Orthogonalization.

This paper concerns the modeling of covariance stationary multiple time series by the method of successive orthogonalization. Fundamental to the method of successive orthogonalization is the factorization of the spectral density matrix which results into a unique triangular two sided moving average model TTSMA model. The proposed method of modeling multiple time series is to build specify, estimate and check a TTSMA model for a given set of observations. Any other form of the model, say the canonical model, can be obtained by an appropriate transformation.