Temporary change detection on ARMA(1,1) data

The aim of this research is to carry out appropriate method to detect Temporary Change on Autoregressive Moving Average (ARMA) (1,1) data. Estimation of model parameters and outlier effects are used to iteratively for joint estimation procedure. Simulation data were generated from ARMA (1,1) model. The ARMA consists of 4 models which were produced by parameters combination of Autoregressive (AR) and Moving Average (MA). Residuals were estimated by Conditional Least Square (CLS) and Median Absolute Deviation (MAD). Removing outliers’ effect used two ways: replacing data which containing outlier and omitting. The observation contains outlier replaces by other value, namely replacement procedure and omitting the observation contains the outlier, namely omit one procedure. The result shows omit one procedure detect outliers better than replacement procedure for all cases. Moreover, MAD and Omit one combination is slightly better than CLS and Omit one combination. This method was implemented to Surabaya’s Air Pollutant (Sulfur Trioxide) data and produced similar result. Joint Estimation method using combination MAD and omit one procedure obtain more accurate to detect Temporary Change than three others procedures.