Identification of the time series interrelationships with reference to dynamic regression models
In this study, the model of interest is that of a rational distributed lag function Y on X plus an independent Autoregressive Moving Average (ARMA) model. To investigate the model structure relating X and Y we considered the inverse cross correlation function for the observed and residual series in the presence of outliers. A two stage identification procedure is presented which involves fitting univariate time series model to each series and identifying a dynamic shock model relating the two univariate model series. The models so far obtained were combined to identify a dynamic regression model, which were fitted in the usual ways. From our findings, there was a reduction in the error variance of the final model with the outlier free stationary series which is an indication that the two-stage procedure is reliable and efficient.
JONAMP Vol. 11 2007: pp. 621-626