Robustness of the rescaled range R/S in the measurement of noncyclic long run statistical dependence