American Journal of Applied Sciences

Robust Outlier Detection in Linear Regression

Nethal K. Jajo and Xizhi Wu

DOI : 10.3844/ajassp.2004.136.148

American Journal of Applied Sciences

Volume 1, Issue 2

Pages 136-148

Abstract

New methodology of robust outlier detection based on Robustly Studentized Robust Residuals (RSRR) examination is well established in linear regression analysis. Two new robust location estimators of linear regression parameters are developed in simple and multiple cases. Based on these robust estimators we obtain RSRR. We used RSRR to derive a new measure of distance to be used in outlier detection. A graphical display using new measure of distance is constructed for detecting multiple outliers. This graphical display provides a distinguish between detected outlier observations and hidden influential (non-outlier) observations. Real data example and simulation technique were used for illustration and confidential. 1991 Mathematics subject classification (Amer. Math. Soc.), Primary 62J20; Secondary 62G35.

Copyright

© 2004 Nethal K. Jajo and Xizhi Wu. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.