Comment by btn

And a Kruskal-Wallis test instead of an ANOVA, so no "magic" would be required.