Table 1.

Main results of the fitted models: In non-balanced models, the overall accuracy is reported as the highest accuracy reached (assessed across all thresholds providing true positive rate above 50% and false positive rate below 50%); for the position of the points on the ROC curves see Figures 2, 3.

	Area under the curve		Overall accuracy for the chosen threshold
	Before	After	Before (%)	After (%)
Mean across the channels	0.7246	0.7425	76.87	70.15
Mean across the channels; balanced	0.7257	0.7257	69.14	72.01
All channels	0.8146	0.8652	77.61	79.85
All channels; balanced	0.8542	0.8941	79.09	83.45
All channels; leave-one-out	0.6420	0.7236	66.42	68.66
All channels; balanced; leave-one-out	0.5942	0.6481	61.47	66.55

For balanced models, the average of overall accuracies across 100 subsamples is reported.