ABSTRACT
To motivate multiple comparisons with the best (MCB), consider the following (somewhat morbid) example. Suppose, among five treatments being compared, two treatments are so bad that given either one most patients die within a short period of time. Then it is possibly not of primary interest to know which of those two treatments is worse; the inference that neither is best suffices. Suppose the second best treatment (among the three remaining treatments) is almost as good as the true best treatment. Then statistical inference identifying both as practically the best may be of interest, for there may be other considerations (e.g. cost, toxicity) impacting on the choice of the treatment. Thus, in these situations not all pair wise comparisons are of interest. The question is then 'What comparisons are of primary interest?'
