![]() | Only 14 pages are availabe for public view |
Abstract Principal component time series are useful to reduce the dimension of a multiple time series. The original series are filtered with a filter that is computed from the Eigen vectors of the spectral density matrix. The original series is taken here as a vector autoregressive moving average (VARMA) time series and a vector bilinear time series. The variance covariance matrix and the spectral Matrix for the general V ARMA series is derived. The bivariate Bilinear time series models are studied in details. Their properties such as existence of stationary solution, ergodicity and inevertibility are studied. The second order moments are derived and therefore the spectral matrix is obtained. The principal component time series of the bivariate models are obtained. A bivariate, or a multiple, bilinear process can be written as a one dimensional process. |