Table 5. Performance of each model possibility.
| Markers included in the model | Performance Indexes | |||
|---|---|---|---|---|
| AIC | BIC | MSE a | AUC a | |
| CAL | 135.39 | 146.77 | 0.176 (0.115–0.252) | 0.816 (0.671–0.949) |
| CD26 | 155.27 | 166.65 | 0.214 (0.158–0.286) | 0.732 (0.567–0.886) |
| EGF | 150.25 | 161.63 | 0.205 (0.152–0.277) | 0.759 (0.606–0.893) |
| CAL+ sCD26 | 133.43 | 147.65 | 0.173 (0.108–0.249) | 0.823 (0.680–0.958) |
| CAL+ EGF | 133.77 | 147.99 | 0.173 (0.110–0.252) | 0.824 (0.677–0.954) |
| CD26+EGF | 146.77 | 160.99 | 0.196 (0.138–0.274) | 0.778 (0.626–0.917) |
| CAL+ sCD26+EGF | 132.64 | 149.70 | 0.169 (0.103–0.247) | 0.828 (0.683–0.962) |
a Area under the ROC curve (AUC) and mean squared error (MSE) based on out-of-bag (OOB) predictions (models are fitted in the training sets with the 75% of the cases and are subsequently used for predicting the group membership of the 25% test cases). Mean AUC and MSE and 95% bootstrap confidence intervals are provided.