Extending Lawson's Algorithm to Include the Huber M-Estimator
Abstract
When fitting a curve to experimental data, there is no guarantee that the data obtained are as accurate as might be expected. The effect of outside influences may cause the data set to contain outliers. These outliers can have a significant effect on any curve which is fitted to such data. The l infinity-norm, which is particularly appropriate for fitting data with uniformly distributed errors, is extremely sensitive to such outliers, since it minimises the maximum error from the data to the curve. Therefore, a technique which approximates a data set using the l infinity-norm, without being adversely affected by outliers, would be a useful addition to the array of tools available. We present numerical examples to illustrate the use of such a technique and also some practical applications to justify its use.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 2000
- Accession Number
- ADP011967
Entities
People
- Colin Ross
- Iain J. Anderson
- John C. Mason
Organizations
- University of Huddersfield