A technical note on the power determination for Fisher's combination test

In a recent paper, BANIK, KOHNE, and BAUER (1996) investigated the use of Fisher's combination test for two stage sampling in terms of its power as compared to the optimal test in the pooled sample. The distributional assumptions corresponded to the application of paired comparisons and parallel groups t-tests in the classical nonadaptive scenario. Allowing early stopping in favor of H-0, as described in BAUER and KOHNE (1994), modifies this test procedure and possibly reduces the sample size. For power calculations they noted a technical problem in integration arising from function singularities. This note provides a technically feasible solution for power calculations in these and other cases using SAS/LML software. Some results, e.g. concerning a potential gain in power when allowing for early stopping are presented.