Journal of Official Statistics, Vol.4, No.2, 1988. pp. 113124

Current Issue
Personal Reference Library (PRL)
Personal Page

On Autoregressive Model Identification

Since Cleveland (1972) introduced the inverse autocorrelation function, it has been recognized as a competitor to the partial autocorrelation function as a time series model identification tool. By using simulated and real data, we have demonstrated that neither of these is consistently more powerful than the other for identification of autoregressive (AR) models. However when the underlying AR process is of full order, the partial autocorrelation function invariably is the superior. But when a subset order AR model generates the data, the inverse autocorrelation function is generally more informative. On the whole the partial autocorrelation function exhibits better performance. For instance, in two of the three cases of real series used it clearly outperforms the inverse autocorrelation function.

Autoregressive model identification: partial autocorrelation function; inverse autocorrelation function.

Copyright Statistics Sweden 1996-2018.  Open Access
ISSN 0282-423X
Created and Maintained by OKS Group