I am applying statistical tests from paper "Statistical Comparisons of Classifiers over Multiple Data Sets" J. Demsar (link) in my own work. However, variance is pretty important, since different classifiers that I am comparing vary wildly in this regard.
Is there any alternative to Friedman-Nemenyi posthoc test that can take variance into consideration? Basically, has anything changed for section 3.2.3 of that paper since 2006?
