A Factor Analysis for Time Series.
Daniel Peña, Box,G E P · Defense Technical Information Center (DTIC) · 1984
This paper studies how to identify hidden factors in multivariate time series process. It is shown that the number of factors must be equal to the rank of both the covariance matrices and the parameter matrices of the infinite moving average representation of the process. A canonical transformation is derived which can recover such factors. The method is illustrated with several examples. Originator-supplied keywords include: Multivariate ARMA process, Canonical analysis, Eigenvalues and Eigenvectors.