On a Robust Correlation Coefficient

Abstract

In this paper the robustness of some well known correlation coefficients, namely, Pearson's, Spearman's and Kendall's, are examined. The empirical evidence shows that these correlation coefficients are sufficiently robust against a substantial number of outliers. That is, they do not have high breakdown points. As an alternative, a correlation coefficient based on the least median of squares is proposed. It is shown that this correlation coefficient has a higher breakdown point than the well known correlation coefficients.