The Scale Problem in Robust Regression M-Estimates

David M Rocke, David F. Shanno

Research output: Contribution to journalArticlepeer-review

3 Scopus citations


Regression M-estimates require an auxiliary scale to determine which residuals are so large that the point should be downweighted. This paper investigates several ways of incorporating scale into the estimation process and compares them theoretically and by Monte Carlo. Computational methods are presented that are superior to those now in common use.

Original languageEnglish (US)
Pages (from-to)47-69
Number of pages23
JournalJournal of Statistical Computation and Simulation
Issue number1
StatePublished - Jan 1 1986


  • Biweight estimate
  • Huber-Dutter algorithm. Hubers Proposal 2 estimate
  • iteratively reweighted least squares
  • median absolute deviation. quasi-Newton methods
  • Winsorized variance

ASJC Scopus subject areas

  • Statistics, Probability and Uncertainty
  • Modeling and Simulation
  • Statistics and Probability
  • Applied Mathematics


Dive into the research topics of 'The Scale Problem in Robust Regression M-Estimates'. Together they form a unique fingerprint.

Cite this