Search

In statistics, the coefficient of determination can be used to assess the suitability of a model with the data. If there are outliers in the data, the coefficient of determination obtained by the OLS method is not consistent. The purpose of this study was to compare the coefficient of determination of regression lines obtained by the OLS, the M and the LMS methods as a measure of the suitability model. The result showed that when the data contains no-outlier, the LMS method is as consistent as the OLS and the M methods concerning the coefficient of determinations. When the data contain outliers, the LMS method is more consistent than the OLS and the M methods. This result was based on real data with 9.1% outliers. Dalam statistik, koefisien determinasi dapat digunakan untuk menilai kesesuaian model dengan data. Jika ada outlier pada data, koefisien determinasi yang diperoleh dengan metode OLS tidak konsisten. Tujuan dari penelitian ini adalah untuk membandingkan koefisien determinasi dari garis regresi yang diperoleh melalui metode OLS, M dan metode LMS sebagai ukuran model kesesuaian. Hasil penelitian menunjukkan bahwa ketika data tidak mengandung-outlier, metode LMS adalah konsisten, serupa dengan metode OLS dan metode M terkait dengan koefisien determinasi. Ketika data mengandung outlier, metode LMS lebih konsisten daripada metode OLS dan metode M. Hasil ini berdasarkan ujicoba pada data nyata dengan outlier 9,1%.

The using of OLS method to estimate the regression coefficients in multiple linear regression model presupposed assumption that there is no outlier in the data. Alternatively, robust regression methods can be used. This paper aims to investigate the efficiency of M method and LMS method to estimate the regression coefficients. Besides application data, simulation data generated by MINITAB and SYSTAT package program were used. The investigation shows the LMS method is more efficient than the M method when there is outlier in the data. Otherwise, the M method is more efficient than the LMS method.