Stochastic independence between random measurements is one of the important aspects of many statistical investigations. In this case, Pearson's product moment correlation is the most commonly used test. Since Pearson's correlation is well-known to be quite sensitive to errors in measurements, rank correlations have been recommended as alternatives. Based on numerical evidence, several rank correlations introduced lately have been claimed more resistant than others. In this paper, we address the analytical quantification of this concept. Finally, five rank correlations are compared according to their resistance properties.
展开▼
