ABSTRACT
We touched upon selection and ranking briefly in Section 2.4. In this chapter, we discuss more fully a selection problem to identify the “best” one among k(≥ 2) treatments. The responses from k treatments are assumed independent and normally
distributed having unknown means and a common unknown variance. We discuss some methodologies under this setup for three reasons. First, it is easy to appreciate basic principles in this setup. Second, the fundamental formulation and seminal approaches proposed for this problem have led to far reaching developments. Third, it may be a good thing to walk through some of the preliminaries in a familiar context.
