Iteratively Reweighted Least Squares - Encyclopedia Entry.
Abstract
Iteratively Reweighted least Squares (IRLS) is a computationally attractive method for providing estimated regression coefficient that are relatively unaffected by extreme observations. Definitions and statistical justifications are reviewed, and a numerical example and a multivariate extension are included. This article is to be an entry in The Encyclopedia of Statistical Sciences. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 01, 1982
- Accession Number
- ADA114534
Entities
People
- Donald B. Rubin
Organizations
- University of Wisconsin–Madison