I hate to say it again, but I just posted some critical comments on non-parametric statistics on one of the DNDN boards- since the computer, these are rarely useful. Medians remove effects of outliners but parametric results should always backup non-parametrics.