We consider a linear normal modelY=X[theta]+ewith[theta]verifying a linear restriction and the standard estimators [theta](unrestricted MLE) and[theta]* (restricted MLE). We prove that[theta]* is preferable to [theta]using a new and strong criterion which implies the domination under other usual...

This paper shows how in a normal population, when the mean holds restrictions given by a right cone, the likelihood ratio test (LRT) for testing a face of that cone is equivalent to another test which is also the LRT for testing the linear subspace associated to the above face against a right...