FIXED-WIDTH SIMULTANEOUS CONFIDENCE-INTERVALS FOR ALL-PAIRWISE COMPARISONS

In this paper we consider the problem of constructing a set of fixed-width simultaneous confidence intervals for all-pairwise differences of the means of k independent normal populations with a common unknown variance. A sequential procedure is necessary to achieve this goal. For the pure sequential procedure considered in this paper, the second order approximations to the confidence level and the expected sample size are given. A recursive method for the exact calculations of the confidence level and the expected sample size is also provided. Comparisons between the second order approximations and the exact calculations are carried out to indicate when the approximations are reasonably accurate.