ROBUST ESTIMATION IN REGRESSION MODEL FOR HANDLING OUTLIER AND HETEROSCEDASTICITY

Abstract

Regression analysis is a method in statistics to determine the relationship between the dependent variable and independent variables. The method used in the estimation of parameters in the model is Ordinary Least Square (OLS). This method is very sensitive to deviations assumptions on the data. The assumption is often not met is assuming heteroscedasticity. One cause of non-fulfillment of this assumption because there are outlier in the data. Therefore, another method used to handle the data outlier. The solve of this case is the robust regression method using estimates Least Trimmed Square (LTS) and Least Median Square (LMS). This paper will be discussed handling outlier and heteroscedasticity by comparing both the robust estimation by OLS seen from the residual standard error, standard error, and the regression coefficient. The results obtained from the Central Java province GRDP data in 2016-2017, show that data containing the direction y outlier, the method of estimating at least the median square is better to use compared to other methods.