Estimate AR(3) by Using Levinson-Durbin Recurrence & Weighted Least Squares Error Methods

Main Article Content

Jinan Abbas Naser

Abstract

In this study, we investigate about the estimation improvement for Autoregressive model of the third order, by using Levinson-Durbin Recurrence (LDR) and Weighted Least Squares Error ( WLSE ).By generating time series from AR(3) model when the error term for AR(3) is normally and Non normally distributed and when the error term has ARCH(q) model with order q=1,2.We used different samples sizes and the results are obtained by using simulation. In general, we concluded that the estimation improvement for Autoregressive model for both estimation methods (LDR&WLSE), would be by increasing sample size, for all distributions which are considered for the error term , except the lognormal distribution. Also we see that the estimation improvement for WSLE method, depends on the value for the Forgetting Factor parameter (α),which haave value less than one(i.e. 1) ( α< ). The estimate is improved for large value for parameterα exactly at 0.99 α= .Finally, we used the estimation methods (LDR&WLSE) for real data.

Article Details

How to Cite
NASER, Jinan Abbas. Estimate AR(3) by Using Levinson-Durbin Recurrence & Weighted Least Squares Error Methods. Ibn AL- Haitham Journal For Pure and Applied Science, [S.l.], v. 26, n. 3, p. 357-378, apr. 2017. ISSN 2521-3407. Available at: <http://jih.uobaghdad.edu.iq/index.php/j/article/view/447>. Date accessed: 25 sep. 2020.
Section
mathematics