Table 4:
Prediction results for testing data with R-CHOP therapy for different methods. The models were estimated using the training data with CHOP therapy. The third column is the number of probe sets selected by a method. The fourth column displays the number of overlapping probe sets in BJ-LS* and other methods. P-value is calculated by a log-rank test for high risk and low risk groups.
| Model | Method | No. | Overlap | P-value |
|---|---|---|---|---|
| AFT linear | BJ-LS* | 12 | - | 0.004 |
| IPC-B* | 57 | 1 | 0.012 | |
| IPC-LASSO | 64 | 1 | 0.011 | |
| BJ-EN | 17 | 12 | <0.001 | |
| AFT non-linear | BJ-Tree (degree=1) | 61 | 8 | <0.001 |
| BJ-Tree (degree=4) | 66 | 8 | <0.001 | |
| BJ-SS* | 23 | 8 | <0.001 | |
| Cox linear | Cox-LASSO | 118 | 12 | 0.016 |
| Superpc | 97 | 8 | <0.001 |