A Comparion of Ordinary Least Squares Regression and Least Squares Ratio via Generated Data

摘要

Regression Analysis (RA) is one of the most used tools for functional relationship. The Ordinary Least Squares (OLS) method is the basic technics of RA. In this study we introduce one of the robust regression approaches, called Least Squares Ratio (LSR), and make a comparison of OLS and LSR according to mean square errors of regression parameter estimations. In this study for certain theoratical model, we generate data for different sample sizes, error variances and number of outliers. It is found that no matter what the sample size is LSR always performs well when there is 1 or more outliers.