A rank-free method of calculating Kendall's τ is described. Derived from Daniels' 1944 general treatment of correlation, it is based on the signs of all possible paired comparisons in a data set. Unlike the 1975 alternative methods of Cooper and those of 1977 by Stuart, it allows for ties and thus yields the same coefficient as that of Kendall's original method. The execution is simpler, however, because it does not require ranking of data.
Get full access to this article
View all access options for this article.
References
1.
CooperM. (1975) A non-parametric test for increasing trend. Educational and Psychological Measurement, 35, 303–306.
2.
DanielsH. E. (1944) The relation between measures of correlation in the universe of sample permutations. Biometrika, 33, 125–135.
3.
KendallM. G. (1938) A new measure of correlation. Biometrika, 30, 81–93.
4.
SchmeidlerJ. (1978) On Cooper's non-parametric test. Educational and Psychological Measurement, 38, 913–915.
5.
StrahanR. F. (1982) Assessing magnitude of effect from rank-order correlation coefficients. Educational and Psychological Measurement, 42, 763–765.
6.
StuartA. (1977) Spearman-like computation of Kendall's tau. British journal of Mathematical and Statistical Psychology, 30, 104–112.
