Perbandingan Metode Ordinary Least Square (OLS) dan Metode Partial Least Square (PLS) Untuk Mengatasi Multikolinearitas

Abstract

Miulticollinearity is a violation of classical assumptions, which is why parameter estimation with the Ordinary Least Squares (OLS) method becomes inefficient but still remains unbiased and consistent. One method that can be used to overcome the limitations of the Ordinary Least Squares (OLS) method is the Partial Least Squares (PLS) method. The purpose of this paper is to compare the efficiency level of the two methods, namely OLS and PLS. This writing uses the literature review method, using 10 readings consisting of 2 books and 8 journal articles to be able to define and analyze the comparison of the level of efficiency of the OLS and PLS methods. The comparison is based on the acquisition of bias value and Mean Square Eror (MSE) value. This analysis shows the OLS method to be an efficient estimator when there is no correlation between independent variables. And PLS has a small bias and MSE value and continues to decrease as the sample size increases. In addition, the PLS method should be used when the independent variables have a correlation of more than or equal to 0.8 (≥/= 0.8).