We consider two classes of risk functions and prove the nonadmissibility resp. admisibility of the usual estimator   for the mean μ of a normal distribution. Furthere, we consider the criteria of G- and D-optimality of experimental designs in connection with choice of a suitable estimator for μ