Simulating Comparisons between Spearman and Kendall’s tau-b Methods for the Nonparametric Bivariant Correlation Analysis

Objective To compare the differences between the nonparametric bivariate correlation analysis of Spearman method and Kendall’s tau-b method by simulating different skew distributions or ordinal variables.Methods Monte Carlo simulating method was used with SAS 9.13 program.Two kinds of skew distributions;three different population correlation coefficients,and four different sample sizes were set up.The sample correlation coefficients and the relative errors were compared between the two nonparametric methods.Results Sample correlation coefficients of the Spearman method were more approximate to population correlation coefficients than that of the Kendall’s tau-b method.Conclusion Spearman method is better than Kendall’s tau-b in nonparametric bivariant correlation analysis for the data with skewness distribution or ordinal variables.

Download(PDF format)