A term (introduced by Wald in 1939) used in statistical inference in several contexts. A procedure is admissible if there is no alternative procedure that performs at least as well under all circumstances and performs better under some circumstances.