Fig. 2

Simulation design. PBO = placebo; TRT = active treatment; ANCOVA = analysis of covariance; GSV = growth scale value (ability score); range = age range condition (3–6 years or 12–16 years); imp = impairment condition (average V-scale for sample of 1, 2, 3, 4, or 5 SD below average); es = effect size condition (zero or large standardized effect). Baseline and follow-up scores were drawn from a bivariate normal distribution, where the within-subject correlation was fixed at r = 0.8 and the mean and standard deviation were defined based on the age range and impairment conditions. For the PBO condition, the mean at baseline and follow-up were identical. For the TRT condition, the baseline mean was identical to PBO, but the follow-up mean was shifted by a factor depending on the effect size condition. Baseline age of the simulated participants was drawn from a uniform distribution, and V-scale scores were obtained using a lookup table with age and the simulated GSV score. Finally, the effect of group on follow-up score was calculated based on ANCOVA, controlling for baseline score. This process was repeated 5000 times (large effect size) or 10,000 times (zero effect size) for each combination of conditions