A Linear Parameter Estimator
by Deborah Sturm.
A method is presented for determining the parameters of linear model functions which performs well in the presence of outliers. Like the M-estimator method, our method, inner products (IP) is a generalization of Least Squares (LS) and yilds robust results. Unlike M-estimators, IP always leads to linear equations which are easy to solve. The IP method consists of multiplying the residuals by weighted coefficients, and then solving the summed equations for the unknown parameters. The weights, which are functions of the data, are chosen so that the influence of outliers is reduced. Results of computer experiments are presented which show IP to be as robust as but computationally simpler than the classical M-estimator methods.
Robust Linear Regression, outliers
Deborah Sturm, email@example.com
READING THE ARTICLE: You can read the article in
portable document (.pdf) format (236452 bytes.)
NOTE: The content of this article is the intellectual property of the authors, who retains all rights to future publication.
This page has been accessed 2546 times since July 24, 2006.
Return to the Home Page.