In that situation, the unequal variances version of the t-check might give extra reliable outcomes. Rank transformations don't preserve variances, however variances are recomputed from samples after r ...