On the admissibility of stable spherical multivariate tests

This paper deals with correlation tests from the class of spherical tests introduced by Lنuter (Biometrics 52 (1996) 964). These methods provide an alternative to classical MANOVA approaches and are particularly useful in small samples. Following a brief introduction of the spherical tests, it is shown that the so-called principal component correlation test is admissible in this class. A Bayesian approach is used to prove this result.