![Sample our Computer Science journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5928/TOC-Subject-Sample-2021-Computer-Science.jpg"})
Abstract
Different techniques for estimation in ARIMA-models are investigated, with special emphasis on the moving-average case. It is shown empirically and theoretically that the backforecasting method suggested by Box and Jenkins (1970) may lead to biased estimates for the MA-parameter(s) as well as for the residual variance. This is particularly so if the series is short and/or the parameters close to the noninvertibility region. An explanation for the numerical problems often encountered in this estimation problem is given too. The results are of special importance when seasonal models are onsidered.