Given two groups of subjects, determine if the the medians (NOT means)
of the two subject are significantly different at each node.
This test does not use the standard Wilcoxon Rank-Sum procedure.
Instead, the data are converted to ranks and fed into the two-sample
T-Test. This implementation is described in Rank
Transformations as a Bridge Between Parametric and Nonparametric
Statistics by W. J. Conover and Ronald L. Iman from The American Statistician, Vol 35,
No 3 (Aug., 1981), pp. 124-129.