Robust and Consistent Estimates of Autoregressive-Moving Average Parameters

Abstract

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.. Biometrika Trust is collaborating with JSTOR to digitize, preserve and extend access to Biometrika. SUMMARYWe aim at estimates of autoregressive-moving average parameters which are simultaneously consistent at the nominal Gaussian model and insensitive to outliers in the data. The estimates may be viewed as a robust version of maximum likelihood estimates of order r and, for purely autoregressive models, reduce to a special case of generalized M-estimates.Some key words: Autoregressive-moving average model; Generalized M-estimate; Maximum likelihood estimation of order r; Robust estimation.

ARFIMA processes and outliers: a weighted likelihood approach

“…Since the weighted log-likelihood estimating equations (Equation (8)) are defined on residuals it will be robust against the presence of innovation outliers while it may be not robust in the presence of AO. This behavior is very well-known in classical M-estimation (see, for instance [20]). …”

Robust Identification of Autoregressive Moving Average Models

“…Box and Jenkins, 1976) are very sensitive to outliers and, therefore, not appropriate to deal with situations of the kind just described. In the last few years, several classes of robust estimators of y have been proposed as an alternative to the usual least squares estimators (see the survey article of Yohai, 1985, andMasarotto, 1985). However, less attention has been paid to the tentative specification stage, i.e., to the identification of p and q.…”