Robust Estimators for the Correlation Measure to Resist Outliers in Data

The objective of this research was to propose a composite correlation coefficient to estimate the rank correlation coefficient of two variables. A simulation study was conducted using 228 situations for a bivariate normal distribution to compare the robustness properties of the proposed rank correlation coefficient with three estimators, namely, Spearman's rho, Kendall's tau and Plantagenet's correlation coefficients when the data were contaminated with outliers. In both cases of non-outliers and outliers in the data, it was found that the composite correlation coefficient seemed to be the most robust estimator for all sample sizes, whatever the level of the correlation coefficient.