Ali, Hazlina and Syed Yahaya, Sharipah Soaad (2013) On robust mahalanobis distance issued from minimum vector variance. Far East Journal of Mathematical Sciences (FJMS), 74 (2). pp. 249-268. ISSN 0972-0871

PDF Restricted to Registered users only Download (42kB) | Request a copy

Abstract

Detecting outliers in high dimension datasets remains a challenging task.Under this circumstance, robust location and scale estimators are usually proposed in place of the classical estimators. Recently, a new robust estimator for multivariate data known as minimum variance vector (MVV) was introduced. Besides inheriting the nice properties of the famous MCD estimator, MVV is computationally more efficient. This paper proposes MVV to detect outliers via Mahalanobis squared distance (MSD).The results revealed that MVV is more effective in detecting outliers and in controlling Type I error compared with MCD.

Item Type:	Article
Uncontrolled Keywords:	robust estimation, Mahalanobis distance, minimum vector variance, outliers detection.
Subjects:	Q Science > QA Mathematics
Divisions:	School of Quantitative Sciences
Depositing User:	Dr. Hazlina Ali
Date Deposited:	16 Apr 2017 02:23
Last Modified:	16 Apr 2017 02:23
URI:	https://repo.uum.edu.my/id/eprint/21569

Actions (login required)

View Item