Power of comparing a number of different data sets with a varying number of dimensions (“Dim”) for which there is an expected value difference δ in the tested data sets.

For the Hall test, the T and S tests gave very similar results. Three nearest neighbors were used in the Nearest Neighbor test. The results of the Friedman-Rafsky test are for three trees, which consistently performed better than one and two trees in this setting. The scale-space results are for the Bonferroni/FDR correction, respectively. A 0.10 significance level is used and 2000 Monte Carlo samples are used.