A Case against Continuing Use of the Spearman Formula for Rank-Order Correlation

Historical development of the Pearson correlation and the Spearman analog on the ranks was traced to delineate the basis on which the Spearman formula was derived. Following this, recommendations from current statistical textbooks on the use of the Spearman formula were presented. Arguments of computational simplicity, appropriateness when data are not at the interval level, and appropriateness when the relation between X and Y is monotonic but nonlinear for using Spearman were found deficient. Results from a small simulation study also supported this view.