Perbandingan metode regresi kuadrat terkecil dengan metode kekar

Abstract

Ordinary least squares method is generally used to estimate parameter of a linear model. However, ordinary least squares is vulnerable against outliers because this method treats every element of data with equal weight. To overcome this problem, several studies have been done to find methods that produce more robust estimator. The robust methods being studied in this paper are weighted least squares, least absolute deviations, least median squares, and least trimmed squares methods. The comparison among those methods indicates that the method of least absolute deviation and weighted least squares are more resistant in case of the presence of outliers in the dependent variable. Meanwhile, least median squares and least trimmed squares methods are resistant to the presence of outliers both in the dependent and independent variable.