Back

ABSTRACT

A Comparison of OLS and Robust Regression using S-PLUS

Researchers need to consider robust estimation methods when analyzing data in multiple regression. The ordinary least squares estimation of regression weights in multiple regression is affected by outliers, non-normality, multicollinearity, and missing data. It is therefore important to check the accuracy and stability of estimates using robust estimation methods. The ordinary least squares, least-trimmed, and MM parameter estimation methods were compared in the present study, which reported widely different results between ordinary least squares and the robust estimation methods. An example was given using The High School and Beyond data set with the S-PLUS statistical package. Textbooks articulate many different robust regression methods, however, only a few software packages permit robust regression estimation.

Download Full Article