This paper aims to compare the relative efficiency of weighted least square (WLS), ordinary least square (OLS) and robust regression method in regression coefficient estimation when the error term is not homogen. The assumption of homegeneous error variance underlying the ordinary least square (OLS) is very important to get the best linear unbiased estimation of the regression coefficients. The investigation compares the methods in calculating efficiency of booth simulation and experimental data. In conclusion, the WLS method is relatively more efficient than OLS and Robust Regression methods.
Copyrights © 2005