Abstract:
The existence of outliers in time series may have a pernicious effect on the estimation of economic and financial signals. Structural changes caused by outliers may reduce the estimated time series model's accuracy and result in forecast failure. The procedure for detecting outliers has been the most crucial issue in this study. We apply a general-to-specific modelling to detect the outlier via indicator saturation in the local level model framework using gets a package embodied in R programming language. Focusing on impulse indicator saturation, we assess its performance by using Monte Carlo simulations. The Monte Carlo experiments revealed that the effectiveness of impulse indicator saturation relies heavily on the size of additive outliers, level of significance, and locations of an outlier in the series. Furthermore, we apply impulse indicator saturations to the detection of outliers in FTSE Bursa Malaysia Hijrah Shariah and FTSE All-World Shariah stock indices.