Abstract
A new criterion for the rejection of outlying observations is derived.