Multiple time series model identification using concatenated sample cross-correlations

Abstract

A weakness of the ESCC approach (Tiao and Tsay, 1983) for model identification of multiple time series is the complication involved in checking all the elements in the ESCC matrix table, which prevents the automation of the method.

In this paper, a new approach is developed for automatic model specification of the vector ARMA model, using the asymptotic property of the concatenated sample cross-correlation (CSCC) matrix. The new approach is illustrated by examples with generated and real data.

Keywords:

• multivariate time series
• vector autogressive moving-average model
• augmented sample cross
• correlation matrix: model identification