A review of an optimization problem (Q1105296)

In this paper we review the theoretical developments as well as the diverse applications reported over a period of 65 years of a basic optimization problem in statistics. The optimization problem considered is a fairly simple one to state. Let X be a normal variate with mean \(\mu\) and standard deviation \(\sigma\) and also let \(-\infty =x_ 0<x_ 1<x_ 2<...<x_ m<x_{m+1}=\infty\). Further let \(p_ i=P(x_{i-1}<X<x_ i)\) and \(\mu_ i=E(X| x_{i- 1}<X<x_ i)\) for \(i=1\) to \(m+1\). The problem under consideration is to find \(x_ 1,x_ 2,...,x_ k\) so that \(\sum^{m+1}_{i+1}p_ i(\mu_ i-\mu)^2\) is maximized. This problem has been encountered in diverse areas such as survey sampling, statistical inference, statistical quality control, discriminant analysis and psychometry.