The complexity of computing the MCD-estimator

Authors:
Thorsten Bernholt;Paul Fischer
Affiliations:
Lehrstuhl Informatik 2, Universität Dortmund, Baroper Str. 301, 44221 Dortmund, Germany;IMM, Technical University of Denmark, Kongens Lyngby, Denmark
Venue:
Theoretical Computer Science
Year:
2004

Citing 5
Cited 6

Matrix analysis

Matrix analysis
The feasible solution algorithm for the minimum covariance determinant estimator in multivariate data

Computational Statistics & Data Analysis
Improved feasible solution algorithms for high breakdown estimation

Computational Statistics & Data Analysis
A fast algorithm for the minimum covariance determinant estimator

Technometrics
The largest nonindentifiable outlier: a comparison of multivariate simultaneous outlier identification rules

Computational Statistics & Data Analysis

Using combinatorial optimization in model-based trimmed clustering with cardinality constraints

Computational Statistics & Data Analysis
RelaxMCD: Smooth optimisation for the Minimum Covariance Determinant estimator

Computational Statistics & Data Analysis
Robust online signal extraction from multivariate time series

Computational Statistics & Data Analysis
Robust data clustering by learning multi-metric Lq-norm distances

Expert Systems with Applications: An International Journal
The Gaussian rank correlation estimator: robustness properties

Statistics and Computing
Using robust dispersion estimation in support vector machines

Pattern Recognition

Quantified Score

Hi-index	5.23

Visualization

Abstract

In modern statistics the robust estimation of parameters is a central problem, i.e., an estimation that is not or only slightly affected by outliers in the data. The minimum covariance determinant (MCD) estimator (J. Amer. Statist. Assoc. 79 (1984) 871) is probably one of the most important robust estimators of location and scatter. The complexity of computing the MCD, however, was unknown and generally thought to be exponential even if the dimensionality of the data is fixed.Here we present a polynomial time algorithm for MCD for fixed dimension of the data. In contrast we show that computing the MCD-estimator is NP-hard if the dimension varies.