Figure 3.

The ROC curves of multivariate logistic regression. Differences between men and women in relative beta power across electrodes; before as well as after the treatment. (A) The model was fitted on the whole dataset using the relative beta power across all 19 electrodes. (B) The model was fitted 100-times on a balanced random subset of 80 patients, and the resulting ROC-curves were averaged. (C) The model was fitted and evaluated using leave-one-out validation scheme. (D) The average of ROC created by using random sex-balanced subsets of the data and using leave-one-out scheme in fitting and evaluating the logistic regression. The black pointers in (A,C) indicate the position on the ROC curve, for which the overall accuracy is reported in the Table 1.