Figure 1.
Maximum type 1 error rate when always selecting the treatment with the maximum effect at interim for an increasing number of treatment groups k. Results are given when using the uncorrected critical boundary
(A) or the fixed Dunnett critical boundary
(B) for equal (black lines) and flexible (gray lines) second-to-first-stage ratios. Nominal one-sided α was set to 0.01 (solid lines), 0.025 (dashed lines), and 0.05 (dotted lines).