While the whole CPM protocol allowed for testing whether binge drinking can be predicted by connectivity edges, Step 1 from CPM allowed us to examine consistent patterns of this connectivity-binge drinking association (consistency analysis). Specifically, the leave-one-out cross-validation created small perturbations of the sample, as there were 119 folds of cross-validation, and in each fold, 1 participant was excluded from the training set. Thus, the correlation between binge drinking and change in connectivity edges was calculated 119 times, with 1 participant being excluded each time. Edge connectivity changes were defined as showing a consistent pattern of association with longitudinal bingeing if they significantly correlated with bingeing in all 119 cross-validation folds.
