An Improved Sequentially Rejective Bonferroni Test Procedure

complexity. These procedures can be used whenever the observed levels of significance are available for all individual tests. It is shown that both the Holm and Shaffer procedures can be improved under the assumption of positive orthant dependence for the test statistics. It is noted that this assumption is met in many important practical situations and recommended that in these cases the new procedure be used in place of its predecessors whenever the required observed significance levels are available. The methodology is illustrated with a numerical example.