Table 3.
Random Forest analysis for the four groups: control subjects, FEP patients, SCZ patients, and ASD patients.
| A. Random Forest statistics | ||||
|---|---|---|---|---|
| Control | FEP | SCZ | ASD | |
| Sensitivity | 92% | 100% | 38% | 17% |
| Specificity | 72% | 85% | 91% | 100% |
| Pos. Pred. Value | 69% | 50% | 60% | 100% |
| Neg. Pred. Value | 93% | 100% | 80% | 83% |
| Balanced accuracy | 82% | 92% | 64% | 58% |
Random Forest statistics over the four groups: Good classification (92% sensitivity, 72% specificity) for control subjects and for FEP patients (100% sensitivity, 85% specificity). ASD and SCZ groups showed high specificity (>90%) but low sensitivity (<38%).
Confusion Matrix of the four groups (after training on 70% of the group populations, N = 59): Correct classification in the validation subgroup (N = 30) of 11/12 control subject, 4/4 FEP patients, 3/8 SCZ patients, and 1/6 ASD patients. Thus, ASD patients were most difficult to classify/identify. See Supplementary Figure S6 for importance of the selected variables.