Abstract
Bechhofer and Turnbull (1978) considered the problems of selecting the best normal population, provided it is better than a standard for the case of known variances or equal but unknown variances. Wilcox (1984) considered the same selection goal for the case of unequal unknown variance and provided the appropriate probability equations and the necessary table. Under the same selection formulation (which we will describe formally in the following sections), this article studies a class of composite inverse sampling procedures for selecting the best multinomial cell that is better than a control cell with unknown cell probability. The procedures guarantee that (a) with probability at least P*0 (specified), no cell is selected when the largest cell probability is sufficiently less than the control and (b), with probability at least P*1 (specified), the cell with the largest probability is selected when its probability is sufficiently greater than the second largest cell probability and the control cell probability.
| Original language | English (US) |
|---|---|
| Pages (from-to) | 985-992 |
| Number of pages | 8 |
| Journal | Biometrical Journal |
| Volume | 30 |
| Issue number | 8 |
| DOIs | |
| State | Published - 1988 |
Keywords
- Composite inverse sampling procedure
- Ranking and selection
ASJC Scopus subject areas
- Statistics and Probability
- Statistics, Probability and Uncertainty