Multi-site classification of OCD patients versus HC was assessed using different cross-validation (CV) approaches. First, we assessed multi-site classification using 10-fold site-stratified CV to obtain maximally homogeneous train-test splits, with approximately the same number of subjects in each fold and the same proportion of samples coming from each site (also referred to as ‘internal validation’). Next, we addressed leave-one-site-out (LOSO) CV, in which all but one site were used to train the models while the left out site was used to assess model performance (external validation). This procedure is then repeated so that each site is used once as a test set. LOSO-CV may result in large between-sample heterogeneity of training and test sets, resulting in lower classification performance34. Because LOSO-CV has different fold sizes, we additionally performed site-stratified CV with varying fold sizes, in which the number of CV folds and respective test-fold sizes are set to match those of LOSO-CV. This was done to evaluate whether differences between site-stratified and LOSO-CV performance were due to differences in heterogeneity or test-fold size variance. Finally, we also performed single-site predictions
