I've always had a preference for tau, because it's got an intuitive interpretation. 

It's not easy to locate literature comparing tau and rho. From what I have read, rho is badly affected by small numbers of extreme outliers. And way back in the days of the 1970 SPSS manual, Nie Hull and Jenkins stated that tau was "more meaningful" when the data contained a large number of tied ranks"

Are there any situations in which Spearman's rho is preferable to Kendall's tau? 

More Ronán Michael Conroy's questions See All
Similar questions and discussions