The Bias Correction of Multiple Linear Regression Parameter Estimators with Residual Bootstrap Method
Proceedings of International Conference on Innovation in Education, Science, and Culture (ICIESC-2017), Medan, Indonesia, 9-10 Nov 2017
7 Pages Posted: 28 Aug 2018 Last revised: 16 Sep 2018
Date Written: November 9, 2017
Abstract
Statistical analysis aims to analyze the relationship between dependent variable with two or more independent variables in the linear parameters model called multiple linear regression analysis. One method can be used to estimate parameters of multiple linear regression using Ordinary Least Squares (OLS). The method of OLS assumes the sample must be normally distributed, where if the sample has outlier so the method of OLS gives parameter estimators of OLS to be biased. The bias that occur can be reduced by using residual bootstrap. The residual bootstrap works by resampling on based the residual obtained from the sample. The outliers on the sample can not be removed of the sample because it can damage the sample characteristics. The simulation results show for a relatively large bootstrap sample giving parameter estimators based on the Bootstrap Ordinary Least Squares (BOLS) more closely to the actual parameters.
Keywords: Regression, Ordinary Least Squares, Residual, Bias, Bootstrap
JEL Classification: C13;C14;C15;C20
Suggested Citation: Suggested Citation