A note on recent proposals for computing l1 estimates

Gilbert W. Bassett, Roger W. Koenker

Research output: Contribution to journalArticlepeer-review

Abstract

Recent methods for computing the least absolute value (l1) estimate have been proposed. In contrast to the usual linear programming formulation of the l1 problem, the new methods attempt to use least squares residuals to identify observations whose l1 residuals are equal to zero. Examples are presented to show that the proposals do not produce estimates that are identical or even necessarily close to l1, and hence the algorithms cannot be recommended as a method for computing the l1 estimate.

Original languageEnglish (US)
Pages (from-to)207-211
Number of pages5
JournalComputational Statistics and Data Analysis
Volume14
Issue number2
DOIs
StatePublished - Aug 1992

Keywords

  • Least absolute value
  • Least squares
  • Linear programming

ASJC Scopus subject areas

  • Statistics and Probability
  • Computational Mathematics
  • Computational Theory and Mathematics
  • Applied Mathematics

Fingerprint Dive into the research topics of 'A note on recent proposals for computing l<sub>1</sub> estimates'. Together they form a unique fingerprint.

Cite this