Fig. 3
From: Machine learning derived risk prediction of anorexia nervosa
Relationship between training dataset with more samples and AUC in the model testing dataset, when moving samples from fold3 to fold2. 10Â % of fold2 corresponds to 129 cases and 312 controls. Error bars with one standard deviation of 10 reruns are shown. Blue horizontal dashed line indicates AUC of fold3 when 100Â % of the fold2 data were employed to train the LR model