Finally, while the consistency analysis revealed some interesting findings, the CPM protocol failed to build models that can predict longitudinal standard/extreme bingeing from change in connectivity edges with leave-one-out cross-validation. Model accuracy was the correlation coefficient between the predicted and observed bingeing in the test set across all cross-validation folds (Rpos: coefficient for the model with edges positively correlated with bingeing, Rneg: coefficient for the model with edges negatively correlated with bingeing). Model accuracy on test sets: standard bingeing Rpos=−0.01 (p=0.48), Rneg=−0.13 (p=.82); extreme bingeing Rpos=−0.04 (p=.59), Rneg=−0.13 (p=.83), with the p-values representing the significant testing results from permutation tests (permuting the data 1000 times).
