Extensions of a monte-carlo comparison of some properties of two rank correlation coefficients in small samples

Spearman Rank Correlation

“…which, for large n, is Whether r S orτ is preferable depends upon the criteria employed to make the judgment; Chow et al [3] judged r S to be the preferable estimator. Spearman's [19,20] introduction of correlation between ranks was accompanied by a correlation measure of which he was fond, the "Spearman footrule", based upon…”

Accurate estimation of standard errors when comparing simulated power functions

“…Here, differences greater than two standard errors are deemed significant corresponding to an asymptotic significance level of approximately 0.05. For our second comparison, we let y be the correlation coefficient of a bivariate normal distribution and we consider tests of significance based on the Kendall rank correlation coefficient, K, and the Spearman rank correlation coefficient, S. Chow, Miller and Dickinson (1974) compared these power functions but were unable to determine which procedure is preferable. To facilitate such comparisons, we performed randomized tests (cr=0.05) and use the more accurate estimate of standard error, SE(C).…”