Outlier Detection in Time Series Model
​
Nurul Sima Mohamad Shariff, Nor Aishah Hamzah, and Karmila Hanim Kamil
​
​
Abstract. Difficulty occurs in time series when the series are contaminated with outliers typically (i) Innovational Outlier (IO) and (ii) Additive Outlier (AO). As such, before estimating the parameters, one needs to overcome the effect of outliers. There are two approaches employed in this study to identify outliers: (i) iterative outlier detection and joint parameter estimates proposed by Chen and Liu [2] and (ii) application of regression diagnostic tools. A simulation study is performed in an effort to assess the performance of both methods. The identification based on the regression diagnostic tools is seems superior compared to those proposed by Chen & Liu. The results also indicate that the proposed technique based on the regression diagnostic tools can be used to determine the outlier effects and the identification on the type of outlier. Moreover, it can also be applied to more complicated time series models that are widely use in practice particularly in the area of statistics research.