Optimal Estimation of Autoregressive Models with Non-stationary Innovations: A simulation study

by K.H. Ng and M.S. Peiris .

Abstract: In many applications of time series, the assumption of stationarity has been widely used to analyse and model time series data. However, this assumption is too restrictive for many real data sets. Therefore, this paper investigates the properties of the class of first order autoregressive (AR(1)) models with constant coefficients and non-stationary innovations (see [5] for details) and considers the parameter estimation in details. The theory of estimating functions (EF) has been reviewed and applied to obtain the corresponding optimal estimates of the respective parameters. A large scale simulation study is carried out to investigate the performance of the resulting estimates. It is shown that the innovation variance affect the estimation of the AR parameter in some cases.

Key Words: AR models, Non-stationary Innovations, Bootstrap, Optimal estimation, Simulations, Efficiency, Estimating Function, Modelling

K.H. Ng, kokhaur@um.edu.my
M.S. Peiris, shelton@maths.usyd.edu.au

Editor: Aqeel, Mohammed.,mohammedageel@hotmail.com

READING THE ARTICLE: You can read the article in portable document (.pdf) format (183092 bytes.)

NOTE: The content of this article is the intellectual property of the authors, who retains all rights to future publication.

This page has been accessed 2248 times since APRIL 26, 2010.

Return to the InterStat Home Page.