Abstract
The main objective of this letter is to compare two concepts of the power of a test developed in the literature on non-nested hypotheses. Using Monte Carlo techniques it is shown that both definitions of power are asymptotically equivalent, but for small samples they differ in some cases very dramatically.