Being not a bitter fanatic of no-parametric tests, nevertheless, I got sometimes little surprised when I find how parametric tests are taken as so perfect that are considered as the comparison?s golden standard. Consequently, for a lot of people, the former should be used only when the Populations under study are not completely known. By the other hand before the George Box?s ironic phrase "essentially, all models are wrong, but some are useful", anyone take it at literal sense: Perhaps Box wanted to prevent against abuses such that to change models every time an insufficiently sound difference is empirically observed. Could be . . .