| dc.description.abstract | Parameter estimation in simple linear regression analysis can be done with Ordinary Least Square (OLS). This method requires the assumption that error must be Normally distributed. However, in reality error is not always Normally distributed, but has another distribution, for example Exponential. Concequently, parameter estimation by MKT is not optimal. This problem can be solved by applying robust regression. This research applied Least Trimmed Square (LTS) method, Winsorized Least Square (WIN) method, and Theil’s method as a solution. The data used in this study was simulation data. Explanatory variables (x) is generated from the Normal distribution, while the error generated from the Exponential distribution. The results of estimation are evaluated with MSE and Relative MSE. The results showed that LTS was the best method when the error is Exponentially distributed. | en |
