An infinitesimal risk reduction will certainly undermine the
complex task of using any of these estimators. As far as the modified LU, LMMS and the MLE
are concerned, one can use any of these depending on one’s preference. Thus d∗, T or the MLE
are all the best alternatives.