Standard Article

Iteratively Reweighted Least Squares

Statistical and Numerical Computing

  1. Murray Jorgensen

Published Online: 15 SEP 2006

DOI: 10.1002/9780470057339.vai022

Encyclopedia of Environmetrics

Encyclopedia of Environmetrics

How to Cite

Jorgensen, M. 2006. Iteratively Reweighted Least Squares. Encyclopedia of Environmetrics. 3.

Author Information

  1. University of Waikato, New Zealand

Publication History

  1. Published Online: 15 SEP 2006

Abstract

Iteratively reweighted least squares (IRLS) is an algorithm for calculating quantities of statistical interest using weighted least squares calculations iteratively. IRLS algorithms may be simply implemented in most statistical packages with a command language because of their use of standard regression procedures. In addition to the required coefficients IRLS produces weights for each observation that aid in understanding the behavior of the algorithm during the iterations. They are also useful diagnostics for identifying unusual data once convergence has been reached.