Penggunaan Metode Regresi Robust untuk Mencari Selang Kepercayaan Koefisien Garis Regresi jika Ragam Galat Tidak Homogen

Harmi Sugiarti (Universitas Terbuka)
Andi Megawarni (Universitas Terbuka)

Jurnal Matematika Sains dan Teknologi
Vol. 4 No. 2 (2003)

The assumption of homogeneous error varianceunderlying the OLS method is very important to get the bestlinear unbiased estimation of the regression coefficients. Thispaper aims to compare the width of confidence interval which isresulted by the OLS, the WLS, and the robust regression methodswhen the error term is not homogen. Comparing the three methodsindicate that the width of confidence interval by the robustregression method is narrower than the OLS method but the WLSmethod is the narrowest.