Fig. 1.

FPR with different versions of the SwE in small samples with Compound Symmetry (ρ = 0.95) and Toeplitz (ψ = 0.1 per visit in the balanced design and ψ = 0.2 per year in the ADNI design) for the balanced and unbalanced ADNI design; all results are based on an F-test at nominal level 5%; S₀, S₁, S₂ and S₃ correspond to the SwE using the raw residuals e_ik, the adjusted residuals ${\mathit{e}}_{\mathit{ik}}\sqrt{\mathit{N}/\left(\mathit{N}-\mathit{p}\right)}$, the adjusted residuals e_ik/(1 − h_ik)^1/2 and the adjusted residuals e_ik/(1 − h_ik), respectively; “Het. with m − p_B”, “Het. with $\widehat{\mathit{\nu }}$” and “Hom with $\widehat{\mathit{\nu }}$” correspond to the standard heterogeneous SwE using (naïvely) m − p_B as degrees of freedom, the standard heterogeneous SwE using the estimate proposed in Eq. (A.16) as degrees of freedom and the modified homogeneous SwE using the estimate proposed in Eq. (A.16) as degrees of freedom, respectively.