Abstract

• This paper aimed to study the effect of outliers in the regression model with a dummy variable based on simulated data. The two robust methods namely Robust Distance Least Absolute Value (RDL1) and Least Trimmed Squares (LTS) and classical method like Ordinary Least Squares (OLS) estimation method were applied to these data. Simulation study showed that the RDL1 and LTS methods detected several outliers whereas the OLS residuals did not reveal any outliers. Based on the mean squared error (MSE) criterion, the RDL1 estimator is more resistant, but it suffers from the swamping effect. The LTS estimator has the second smallest MSE and the OLS method has the largest MSE in this case. The OLS regression is the best when data are free from outliers.