Robustness of the Rescaled Range R/S in the Measurement of Noncyclic Long Run Statistical Dependence

The rescaled range R(t, s) / S(t, s) is shown by extensive computer simulation to be a very robust statistic for testing the presence of noncyclic long run statistical dependence in records and, in cases where such dependence is present, for estimating its intensity. The processes examined in this paper extend to extraordinarily non-Gaussian processes with huge skewness and/or kurtosis (that is, third and/or fourth moments).