Quadratic statistics in testing problems of large dimension
Dmitrii Mikhailovich Chibisov · Lecture notes-monograph series · 2001
We consider testing a simple hypothesis about the mean vector of an iV-variate normal distribution against shift alternatives in a Bayesian setting specifying a prior distribution of the mean vector under the alternative.We treat the problem asymptotically, as N -• oo, and state fairly general conditions on the sequence of prior distributions under which the Bayes tests have asymptotically ellipsoidal acceptance regions.