Accuracy (ACC), AUC, false positive rate (FPR), and false negative rate (FNR) are obtained using supervised (logistic regression) and unsupervised (Gaussian mixture modeling) classification.

All metrics are based on a probability of 0.5 classification threshold. Columns contain tables for clustering using the first canonical correlation variable and first n = 1, 2 principle components from the 9 questionnaires. Raw ordinal scores are used for comparison. Results are grouped into two column groups: the left group contains clustering comparing C vs. CS, the left group shows CC vs. S. Results for logistic regression are the mean of a 10-fold across-validation analysis.