Application of MFV-robustified Correlation Coefficient for the Investigation of the Strength of Beta-convergence of EU NUTS regions

Identification of the strength of relationship between two variables is usually done by the investigation of their correlation coefficient. For determining the strength of linear relationships usually Pearson’s correlation coefficient is used that requires the variables to be normally distributed. However, presence of outliers and even minor deviations in the underlying distribution from the Gaussian can lead researchers and practitioners astray when interpreting the results. Since detection of multidimensional outliers is not always feasible and our data sample does not necessarily shows normal distribution in practice and furthermore correlation analysis is a fundamental step for selecting variables for various statistical analysis a robustified correlation measure is advised to be calculated alongside conventional Pearson’s correlation coefficient. In this study we apply the Most Frequent Value (MFV) robustified correlation coefficient for investigating the speed of convergence in economic growth among NUTS regions of the European Union, where from theory linear relationship is assumed but the corresponding available data are not normally distributed and may contain outliers.