Suppose you want to test whether there is a difference in means in a matched pair, before-and-after a situation. If you know that the populations under investigation are at least mound-shaped and symmetric and you have a large sample, is it better to use the parametric paired-differences test or the nonparametric sign test for matched pairs.