An Anscombe type robust regression statistic

Authors:
Dan Bradu;Douglas M. Hawkins
Affiliations:
Department of Statistics, TwW Bldg 7-126, University of South Africa, P O Box 392, Pretoria 0001, South Africa;Department of Applied Statistics, University of Minnesota, 1994 Buford Avenue, St. Paul, MN 55108, USA
Venue:
Computational Statistics & Data Analysis
Year:
1995

Citing 4
Cited 0

Robust regression and outlier detection

Robust regression and outlier detection
Computing the exact least median of squares estimate and stability diagnostics in multiple linear regression

SIAM Journal on Scientific Computing
The feasible set algorithm for least median of squares regression

Computational Statistics & Data Analysis
Sample size requirements for multiple outlier location techniques based on elemental sets

Computational Statistics & Data Analysis

Quantified Score

Hi-index	0.03

Visualization

Abstract

A new robust regression estimator is proposed. Its use involves sampling of elemental set in a schema very similar to Rousseeuw's least median of squares. Since the construction of such a statistics is done on the basis of residuals from regression, the problem reduces to parameter estimation in a one-dimensional sample, in the face of outliers. Our proposal relies on the work done by Anscombe; it uses the ideas of insurance premium and protection applied to outlier identification, with the addition of Rosner's backwards elimination. This new estimator may represent a modest improvement over methods like the LMS, in that it appears to be able to solve marginal cases resistant so far, requiring the extraction of fewer elemental sets in order to reach a reasonable likelihood of success. The proposed method, however, is not uniformly preferable to the LMS and it should complement the latter rather than replace it.