Fig. 4: Average Classification Confusion Matrix Across All PAF Conditions, ROC curve, and Individual Data Points across All Conditions for all peak alpha.

A) Average classification error confusion matrix showing classification performance in percentages of whether the true class was identified by the NBC. Percentage values reflect the average NBC performance across 2000 iterations of model building and across the 17 participants with data for all conditions. Misclassification is largely due to the overlap of placebo and baseline, but misclassification of drug and placebo can be attributed to carryover effects. B) ROC plot for PAF across head regions. Electrode selections included all electrodes (whole head), all frontal electrodes (frontal), all occipital electrodes (occipital), and all variables together (whole head, frontal, and occipital). Peaking toward the left upper corner indicates better performance (i.e., maximizing true positive rate and minimizing false positive rate). C) Individual data points for PAF across frontal, occipital, and all (i.e., whole head) electrodes.