Computable robust estimation of multivariate location and shape in high dimension using compound estimators

David L. Woodruff, David M Rocke

Research output: Contribution to journalArticle

83 Citations (Scopus)

Abstract

Estimation of multivariate shape and location in a fashion that is robust with respect to outliers and is affine equivariant represents a significant challenge. The use of compound estimators that use a combinatorial estimator such as Rousseeuw’s minimum volume ellipsoid (MVE) or minimum covariance determinant (MCD) to find good starting points for high-efficiency robust estimators such as S estimators has been proposed. In this article we indicate why this scheme will fail in high dimension due to combinatorial explosion in the space that must be searched for the MVE or MCD. We propose a meta-algorithm based on partitioning the data that enables compound estimators to work in high dimension. We show that even when the computational effort is restricted to a linear function of the number of data points, the algorithm results in an estimator with good asymptotic properties. Extensive computational experiments are used to confirm that significant benefits accrue in finite samples as well. We also give empirical results indicating that the MCD is preferred over the MVE for this application.

Original languageEnglish (US)
Pages (from-to)888-896
Number of pages9
JournalJournal of the American Statistical Association
Volume89
Issue number427
DOIs
StatePublished - 1994

Fingerprint

Minimum Covariance Determinant
Minimum Volume Ellipsoid
Robust Estimation
Higher Dimensions
Estimator
S-estimator
Robust Estimators
Computational Experiments
Equivariant
Explosion
Linear Function
Asymptotic Properties
Outlier
High Efficiency
Partitioning
Robust estimation

Keywords

  • Minimum covariance determinant estimator
  • Minimum volume ellipsoid estimator
  • S estimator

ASJC Scopus subject areas

  • Statistics and Probability
  • Statistics, Probability and Uncertainty

Cite this

Computable robust estimation of multivariate location and shape in high dimension using compound estimators. / Woodruff, David L.; Rocke, David M.

In: Journal of the American Statistical Association, Vol. 89, No. 427, 1994, p. 888-896.

Research output: Contribution to journalArticle

@article{80591751fdb74d5fb7bcb58b76ab0e66,
title = "Computable robust estimation of multivariate location and shape in high dimension using compound estimators",
abstract = "Estimation of multivariate shape and location in a fashion that is robust with respect to outliers and is affine equivariant represents a significant challenge. The use of compound estimators that use a combinatorial estimator such as Rousseeuw’s minimum volume ellipsoid (MVE) or minimum covariance determinant (MCD) to find good starting points for high-efficiency robust estimators such as S estimators has been proposed. In this article we indicate why this scheme will fail in high dimension due to combinatorial explosion in the space that must be searched for the MVE or MCD. We propose a meta-algorithm based on partitioning the data that enables compound estimators to work in high dimension. We show that even when the computational effort is restricted to a linear function of the number of data points, the algorithm results in an estimator with good asymptotic properties. Extensive computational experiments are used to confirm that significant benefits accrue in finite samples as well. We also give empirical results indicating that the MCD is preferred over the MVE for this application.",
keywords = "Minimum covariance determinant estimator, Minimum volume ellipsoid estimator, S estimator",
author = "Woodruff, {David L.} and Rocke, {David M}",
year = "1994",
doi = "10.1080/01621459.1994.10476821",
language = "English (US)",
volume = "89",
pages = "888--896",
journal = "Journal of the American Statistical Association",
issn = "0162-1459",
publisher = "Taylor and Francis Ltd.",
number = "427",

}

TY - JOUR

T1 - Computable robust estimation of multivariate location and shape in high dimension using compound estimators

AU - Woodruff, David L.

AU - Rocke, David M

PY - 1994

Y1 - 1994

N2 - Estimation of multivariate shape and location in a fashion that is robust with respect to outliers and is affine equivariant represents a significant challenge. The use of compound estimators that use a combinatorial estimator such as Rousseeuw’s minimum volume ellipsoid (MVE) or minimum covariance determinant (MCD) to find good starting points for high-efficiency robust estimators such as S estimators has been proposed. In this article we indicate why this scheme will fail in high dimension due to combinatorial explosion in the space that must be searched for the MVE or MCD. We propose a meta-algorithm based on partitioning the data that enables compound estimators to work in high dimension. We show that even when the computational effort is restricted to a linear function of the number of data points, the algorithm results in an estimator with good asymptotic properties. Extensive computational experiments are used to confirm that significant benefits accrue in finite samples as well. We also give empirical results indicating that the MCD is preferred over the MVE for this application.

AB - Estimation of multivariate shape and location in a fashion that is robust with respect to outliers and is affine equivariant represents a significant challenge. The use of compound estimators that use a combinatorial estimator such as Rousseeuw’s minimum volume ellipsoid (MVE) or minimum covariance determinant (MCD) to find good starting points for high-efficiency robust estimators such as S estimators has been proposed. In this article we indicate why this scheme will fail in high dimension due to combinatorial explosion in the space that must be searched for the MVE or MCD. We propose a meta-algorithm based on partitioning the data that enables compound estimators to work in high dimension. We show that even when the computational effort is restricted to a linear function of the number of data points, the algorithm results in an estimator with good asymptotic properties. Extensive computational experiments are used to confirm that significant benefits accrue in finite samples as well. We also give empirical results indicating that the MCD is preferred over the MVE for this application.

KW - Minimum covariance determinant estimator

KW - Minimum volume ellipsoid estimator

KW - S estimator

UR - http://www.scopus.com/inward/record.url?scp=21844492457&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=21844492457&partnerID=8YFLogxK

U2 - 10.1080/01621459.1994.10476821

DO - 10.1080/01621459.1994.10476821

M3 - Article

AN - SCOPUS:21844492457

VL - 89

SP - 888

EP - 896

JO - Journal of the American Statistical Association

JF - Journal of the American Statistical Association

SN - 0162-1459

IS - 427

ER -