Detection of Multidimensional Outliers using Biplot Analysis

Authors

  • T. A. Sajesh University of Madras, Chennai
  • M. R. Srinivasan University of Madras, Chennai

DOI:

https://doi.org/10.12723/mjs.13.2

Abstract

It is necessary to examine the valuable data being distorted by the presence of outliers before the same is subjected to necessary analysis. Outliers should be identified using reliable detection methods and tested prior to performing data analysis. Detection of outliers in multidimensional data is important in many applications as it will have far reaching consequences in its analysis. There are methods available in the literature for detecting multiple outliers but there exist no unified method for detecting the same. An attempt has been made to detect the multidimensional outliers through Biplot analysis using elliptical method with a well defined axis (a, b) based on Inter Quartile Range (IQR).The performance of the designed methods is examined by a comparison with the existing methods.

Published

2008-11-30