Alternative Outliers Detection Procedures In Linear Regression Analysis : A Comparative Study


A Common problem in linear regression analysis is outliers, which produces undesirable effects on the least squares estimates. Many widely used regression diagnostics procedures have been introduced to detect these outliers. However, such diagnostics, which are based on the least squares estimates, are not efficient and cannot detect correctly swamping and masking effects. In this paper, we attempt to investigate the robustness of some well known diagnostics tools, namely, Cook’s distance, the Welsch-Kuh distance and the Hadi measure. The robust version of these diagnostics based on the Huber-M estimation have been proposed to identify the outliers. A simulation study is performed to compare the performance of the classical diagnostics with the proposed versions. The findings of this study indicate that, the proposed alternative versions seem to be reasonable well and should be considered as worthy robust alternative to the least squares method

Keywords: Diagnostics, Huber–M Estimation, Least Squares Method, Linear Regression, Outliers, Simulation Study

Unique Article ID: IJMSS-1333

Article Review Status: Published

Pages: 25-33 (Download PDF)

Creative Commons Licence
This work by European American Journals is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License