Advances and Applications in Statistics
Volume 54, Issue 1, Pages 43 - 67
(January 2019) http://dx.doi.org/10.17654/AS054010043 |
|
ROBUST RECURSIVE ANALYSIS OF SEASONAL MOVING AVERAGE MODELS
Mohamed A. Ismail, Hend A. Auda and Mahmoud M. Sadek
|
Abstract: Ismail et al. [10] proposed a robust modifications of the recursive innovative substitution (IS) algorithm of Koreisha and Pukkila [11], for the fitting of moving average time-series model of order (q). They investigated the empirical efficiency and forecasting precision of the two robust modified estimators MIS1 and MIS2, respectively over a variety of situations. In this paper, we consider the work of Ismail et al. [10] of the robust modifications of the recursive IS procedure for fitting seasonal moving average (SMA) time-series model of order (q, Q) under the departure of normality and the existence of outliers effect in the series. A comprehensive Monte Carlo simulation study is presented to compare the performance of the robust modified estimators, MIS1 and MIS2, with both of the IS and the maximum likelihood estimates (MLE) in terms of efficiencies and forecasts precision. The robust modification fits have high asymptotic relative efficiency and better forecasting properties relative to both of the IS and the MLE fits under both symmetric and asymmetric heavy tailed random error distributions and the existence of outliers situations, while losing little efficiency for normal errors situations. |
Keywords and phrases: rank-based method, robust estimation, seasonal moving average, innovative substitution, additive outliers, innovative outliers.
|
|
Number of Downloads: 371 | Number of Views: 3579 |
|