A Parametric Empirical Bayesian Method for Multiple Comparisons

  • Kim, Woo-Chul (Department of Computer Science and Statistics, Seoul National University) ;
  • Hwang, Hyung-Tae (Department of Computer Science and Statistics, Dankook University)
  • Published : 1991.06.01

Abstract

For all pairwise comparisons of treatments, Bayesian simultaneous confidence intervals are proposed and studied. First Bayesian solutions are obtained for a fixed prior, and then prior parameters are estimated by a parametric empirical Bayesian method. The nominal confidence level is shown to be controlled asymptotically. An extension to the unbalanced design is also considered.

Keywords