Table 3.
Summaries of final predictive model sizes (i.e., number of variables used) selected within the clinical-only models and best-performing pipelines, in terms of their mean, median, standard deviation, and coefficient of variation (CoV), as well as the overall number of features selected at least once across 100 bootstraps.
| Mean | Median | SD | CoV | Overall | |
|---|---|---|---|---|---|
| Clinical-only | |||||
| Cox | 4.55 | 5 | 1.250 | 27.5% | 8 |
| LASSO | 3.51 | 3 | 1.541 | 43.9% | 14 |
| Boosted | 5.10 | 5 | 0.870 | 17.1% | 14 |
| RSF | 4.96 | 5 | 1.524 | 30.7% | 8 |
| Top performing pipelines | |||||
| Cox | 9.46 | 10 | 0.926 | 9.8% | 623 |
| LASSO | 11.88 | 12 | 3.891 | 32.8% | 343 |
| Boosted | 14.98 | 15 | 3.052 | 20.4% | 323 |
| RSF | 17.85 | 16 | 9.057 | 50.7 % | 335 |