Abstract
The problem of comparing treatments with a control and with the “best” is considered using a confidence interval approach and selection decisions based on these intervals are discussed, Three cases are considered in detail:
(i):the most sensitive possible comparisons with the control are required.
(i i):the most sensitive possible comparisons with the best are required, and.
(i i i):comparisons with the control and with the best are required to be equally sensitive.
Special results for normal distributions with common unknown variance are obtained and the necessary tables are presented.