Table 2. nQi discrimination performance with different cut-off points in the combined dataset.
Misclassification cost (cost per FN/cost per FP) | Estimated cut-off point | Sensitivity | Specificity | Accuracy | TP | FN | TN | FP |
---|---|---|---|---|---|---|---|---|
1/1 | 0.078 | 0.71 | 0.84 | 0.78 | 30 | 12 | 36 | 7 |
2/1 | 0.075 | 0.74 | 0.79 | 0.76 | 31 | 11 | 34 | 9 |
1/2 | 0.105 | 0.55 | 0.95 | 0.75 | 23 | 19 | 41 | 2 |
TP: true positives, FN: false negatives, TN: true negatives, FP: false positives. The cut-off points have been automatically estimated by maximizing the generalized Youden Index35 under three different misclassification costs assumptions: the cost for FN and FP is equal, the FN misclassification cost is twice the one for FP and that the FP misclassification cost is twice the one for FN.