0.001), and using neuroimaging data following correction for covariates resulted in AUC of 0.52 (95% CI = 0.45–0.59; pcorrected = 1). Finally, classification of medicated vs. unmedicated patients using neuroimaging data only resulted in AUC of 0.74 (95% CI = 0.67–0.81; pcorrected < 0.001), using covariates only in AUC of 0.84 (95% CI = 0.78–0.92; pcorrected < 0.001), and using neuroimaging data after correction for covariates resulted in AUC of 0.59 (95% CI = 0.50–0.67; pcorrected = 0.2). A full overview of these control experiment results can be found in supplementary Tables S8–9. These results show that the covariates contained sufficient information to classify these subgroups, and especially to distinguish between medicated vs. unmedicated patients, as subgroups were already partially defined by site. Nevertheless, the classifications using neuroimaging data only were comparable to the classifications reported above when covariates were added as features. However, correcting the neuroimaging data for covariates lowered the performance substantially compared to adding the covariates as features, suggesting that neuroimaging data were partially related to the covariates (e.g., different scanners at different sites).
