Equality of OLS and Aitken Estimators
- Authors: Belov A.G.1
-
Affiliations:
- Faculty of Computational Mathematics and Cybernetics, Moscow State University
- Issue: Vol 28, No 1 (2017)
- Pages: 74-77
- Section: III. Numerical Methods
- URL: https://journals.rcsi.science/1046-283X/article/view/247564
- DOI: https://doi.org/10.1007/s10598-016-9346-x
- ID: 247564
Cite item
Abstract
We investigate the properties of multiple linear regression estimators obtained by the Ordinary Least Squares method (OLS) and the Generalized Least Squares method (GLS, Aitken estimator) assuming substantially different errors in input data. We illustrate the mechanism that leads to fallacious conclusions about the quality of OLS and Aitken estimators based on visual graphical analysis of experimental data. Numerical simulation and comparative analysis of the estimators is carried out for simple and parabolic regression models.
Keywords
About the authors
A. G. Belov
Faculty of Computational Mathematics and Cybernetics, Moscow State University
Author for correspondence.
Email: ba511@bk.ru
Russian Federation, Moscow
Supplementary files
