Mahalanobis distance and its application for detecting multivariate outliers

Abstract

 While  methods of  detecting  outliers is frequently implemented  by statisticians when analyzing   univariate data,  identifying outliers in multivariate data pose challenges that univariate data do not.   In this paper, after short  reviewing some tools for univariate   outliers  detection,  the Mahalanobis distance,  as a famous
multivariate statistical distances and its ability to detect multivariate outliers  are discussed.  As an application the  univariate and multivariate  outliers of a real data set has been detected using  R  software environment for statistical computing.