Comparison of ordinary and robust harmonic regression in fitting daily mean sea level of west and east peninsular Malaysia

Abstract

Several robust estimations are introduced in hoping to reduce the effect of the unexpected events by providing more reliable estimation and fit. Four east peninsular sea stations and another four west peninsular sea station are studied. Robust M, MM and LTS estimator are being compared each other and result shows that MM has the lowest true outlier and the sum of absolute deviation (SAD), indicating the best fit among the three proposed robust method. The MM estimator is then compared to the OLS, and the results shows that not only MM has the lowest sum of absolute deviation (SAD), but also efficiently ignoring many outliers to redistribute more residuals closer to zero compare to OLS for all stations, indicating MM has the most reliable fit for this study.

Jaffri Nasir

Lecturer & Statistician

My research interests include change-point/threshold estimation, probability theory and optimizations in statistics.