SIMULATION STUDY OF ROBUST REGULATORS PARISETER AND SIDER-S ROBUST PARAMETERS IN LONGITUDINAL DATA WITH A VARIETY OF LEVELS

Robust regression is used to obtain the right model when the data contains outliers and are not normally distributed. Robust regression has several kinds of estimators, one of which is using M-estimator and S-estimator. The M-estimator robust regression is the simplest approach both computationally and theoretically while the S-estimator is an estimator that has a high breakdown point for estimating error scales. This study wants to find out the comparison of M-estimator and S-estimator robust estimation regression that is more efficient by comparing the variance between estimators using relative efficiency in longitudinal simulation data. The results showed that the model with parameter estimators obtained from the S-estimator robust regression method was more effectively used to predict malnutrition in East Java Regency / City in 2013 - 2018 compared to the M-estimator robust regression method.