Performance Analysis of Robust Locations Estimators

Abstract

This paper describes novel computational procedures that reduces the influence of outliers by transformational process. It transforms the influential observations to moderate data points. The procedures utilize all the information contained in the data set. These techniques are compared with the existing robust procedures, such as the weighted mean based on Mahalanobis distance, minimum covariance determinant (MCD), trimmed and winsorized mean. The performance analysis showed that the proposed techniques performed comparably with the well-established robust procedures based on the real data set.