Table 3.

Sample size computation and simulation of empirical power (# replications=10000) for Design 3 where subgroup means: μ_A1 = 15, μ_A2 = 17, μ_A3 = 19, μ_A1A3 = μ_A2A3 = μ_A3A2 = 15, μ_A1A2 = 20, μ_A2A1 = 22, μ_A3A1 = 24; subgroup variances: ${\sigma }_{A_j}^2=6^2,{\sigma }_{A_j{A}_l}^2=8^2$ for j ≠ l; j, l = 1, 2, 3. Response rate for induction treatment A₃ is assumed to be 50%. Hypothesis of interest is H₀ : μ₁₂ = μ₁₃ = μ₂₁ = μ₂₃ = μ₃₁ = μ₃₂.

Scenario	π1	π2	Nominal Power	Overall Sample Size	Empirical Power	Effect Size (Mahalanobis Distance)
1	0.5	0.5	0.8	108	0.83	0.119
	0.2	0.5	0.8	111	0.83	0.116
	0.5	0.5	0.9	139	0.91	0.119
	0.2	0.5	0.9	142	0.91	0.116
2	0.2	0.2	0.8	95	0.84	0.135
	0.2	0.6	0.8	116	0.84	0.110
	0.2	0.2	0.9	122	0.92	0.135
	0.2	0.6	0.9	149	0.91	0.110
3	0.3	0.5	0.8	111	0.83	0.116
	0.3	0.6	0.8	116	0.82	0.110
	0.3	0.5	0.9	142	0.91	0.116
	0.3	0.6	0.9	149	0.92	0.110
4	0.4	0.5	0.8	110	0.83	0.117
	0.4	0.6	0.8	115	0.82	0.111
	0.4	0.5	0.9	141	0.92	0.117
	0.4	0.6	0.9	148	0.91	0.111

Alternative is true with means: Scenario 1: μ₁₂ = 17.5, μ₁₃ = 15.0, μ₂₁ = 19.5, μ₂₃ = 16.0, μ₃₁ = 21.5, μ₃₂ = 17.0

Scenario 2: μ₁₂ = 19.0, μ₁₃ = 15.0, μ₂₁ = 21.0, μ₂₃ = 15.4, μ₃₁ = 21.5, μ₃₂ = 17.0

Scenario 3: μ₁₂ = 18.5, μ₁₃ = 15.0, μ₂₁ = 19.5, μ₂₃ = 16.0, μ₃₁ = 21.5, μ₃₂ = 17.0

Scenario 4: μ₁₂ = 18.0, μ₁₃ = 15.0, μ₂₁ = 19.5, μ₂₃ = 16.0, μ₃₁ = 21.5, μ₃₂ = 17.0