TABLE 7.
Area under the receiver operating characteristic (ROC) curve derived using c-statistic for 10-fold cross-validation sets.
| Proximal | Aerial | |||||
|
|
|
|||||
| Folds | N | Ordinal | Binary | N | Ordinal | Binary |
| 1 | 16 | 0.83 | 0.82 | 16 | 0.92 | 0.94 |
| 2 | 16 | 0.83 | 0.82 | 16 | 0.93 | 0.93 |
| 3 | 16 | 0.83 | 0.81 | 17 | 0.92 | 0.93 |
| 4 | 17 | 0.84 | 0.83 | 17 | 0.93 | 0.94 |
| 5 | 17 | 0.82 | 0.82 | 17 | 0.92 | 0.93 |
| 6 | 17 | 0.83 | 0.81 | 17 | 0.93 | 0.93 |
| 7 | 17 | 0.83 | 0.82 | 17 | 0.92 | 0.94 |
| 8 | 17 | 0.84 | 0.83 | 17 | 0.94 | 0.93 |
| 9 | 17 | 0.83 | 0.85 | 17 | 0.92 | 0.92 |
| 10 | 17 | 0.82 | 0.82 | 17 | 0.92 | 0.93 |
| Total | 167 | 168 | ||||
| Mean | 0.83 | 0.82 | 0.93 | 0.93 | ||
N is the number of data points assigned to the validation set for that fold. The model used for the c-statistic on each validation set was trained using the remaining (167 – N for proximal, and 168 – N for aerial) data points. All 10-fold sets were randomly generated.