On the failures of the null-hypothesis test