Identifying multivariate discordant observations: a computer-intensive approach

Authors:
H. Viljoen;J. H. Venter
Affiliations:
School of Computer, Statistical and Mathematical Sciences, Potchefstroom University, Potchefstroom, South Africa;School of Computer, Statistical and Mathematical Sciences, Potchefstroom University, Potchefstroom, South Africa
Venue:
Computational Statistics & Data Analysis
Year:
2002

Citing 3
Cited 0

Robust regression and outlier detection

Robust regression and outlier detection
Applied multivariate statistical analysis

Applied multivariate statistical analysis
Identifying multiple discordant observations: a computer intensive approach

Computational Statistics & Data Analysis

Quantified Score

Hi-index	0.03

Visualization

Abstract

The problem of identifying multiple outliers in a multivariate normal sample is approached via successive testing using P-values rather than tabled critical values. Caroni and Prescott (Appl. Statist. 41 (1992) 355) proposed a generalization of the EDR-ESD procedure of Rosner (Technometrics 25 (1983)). Venter and Viljoen (Comput. Statist. Data Anal. 29 (1999) 261) introduced a computer intensive method to identify outliers in a univariate outlier situation. We now generalize this method to the multivariate outlier situation and compare this new procedure with that of Caroni and Prescott (Appl. Statist. 41 (1992) 355).