Some Robust Methods of Estimation in Factor Analysis in the presence of outliers

No Thumbnail Available

Date

2010

Journal Title

Journal ISSN

Volume Title

Publisher

International Center for Advance Studies, India

Abstract

A robust method of estimating covariance matrix in multivariate data set is proposed. The goal is to compare the proposed method with the widely used robust methods (Minimum Volume Ellipsoid and Minimum Covariance Determinant) and the classical method (MLE) in the area of Factor Analysis. MVE- MCD- and Proposed- factor analyses use robust covariance matrix in estimating the factor loadings while the classical factor analysis estimates the factor loadings using the MLE of the covariance matrix. While classical factor analysis is found to be more reliable when there are few or no outliers in the data, robust factor analysis will be preferred in the presence of multiple outliers. The Proposed Robust Factor analysis is found to compete favourably well with existing robust methods.

Description

Keywords

Covariance Matrix, Communalities, Factor loadings, Minimum Covariance Determinant, Minimum Volume Ellipsoid, Mean Square Error, Optimality criteria

Citation

ICASTOR Journal of Mathematical Sciences

Collections