Aziz, Nazrina (2014) Eigenstructure-based angle for detecting outliers in multivariate data. Sains Malaysiana, 42 (12). pp. 1973-1977. ISSN 0126-6039

PDF Restricted to Registered users only Download (540kB) | Request a copy

Abstract

There are two main reasons that motivate people to detect outliers; the first is the researchers' intention; see the example of Mr Haldum's cases in Barnett and Lewis. The second is the effect of outliers on analyses. This article does not differentiate between the various justifications for outlier detection.The aim was to advise the analyst about observations that are isolated from the other observations in the data set. In this article, we introduce the eigenstructure based angle for outlier detection.This method is simple and effective in dealing with masking and swamping problems. The method proposed is illustrated and compared with Mahalanobis distance by using several data sets.

Item Type:	Article
Uncontrolled Keywords:	Angle; Eigenstructure; masking; outliers; swamping
Subjects:	Q Science > QA Mathematics
Divisions:	School of Quantitative Sciences
Depositing User:	Dr. Nazrina Aziz
Date Deposited:	20 Dec 2015 08:06
Last Modified:	17 Apr 2016 07:50
URI:	https://repo.uum.edu.my/id/eprint/16543

Actions (login required)

View Item