A SIMPLIFIED GLS ESTIMATOR FOR AUTOREGRESSIVE MOVING-AVERAGE MODELS

Koreisha and Pukkila (1990a) have recently proposed a fast and efficie nt GLS estimator for the univariate ARMA time series model which appea rs to be far more robust than maximum likelihood methods and of compar able accuracy. The one drawback to this new estimator is that it requi res use of the Cholesky decomposition. The purpose of this paper is to suggest an alternative simplified GLS estimator, which can be impleme nted with just repeated applications of an OLS subroutine. A limited M onte Carlo study establishes that this new estimator is just as effici ent as that of Koreisha and Pukkila.