| GLMnet | GBM | Hybrid | ||||||
|---|---|---|---|---|---|---|---|---|
| Study | Phase | Threshold | Sensitivity | PPV | Sensitivity | PPV | Sensitivity | PPV |
| AAP | Original | 0.001 | 1 | 0.078 | 1 | 0.07 | 1 | 0.07 |
| 0.01 | 1 | 0.168 | 1 | 0.138 | 1 | 0.118 | ||
| 0.02 | 0.978 | 0.215 | 1 | 0.274 | 1 | 0.194 | ||
| 0.1 | 0.901 | 0.392 | 0.934 | 0.436 | 0.956 | 0.385 | ||
| Update | 0.001 | 1 | 0.033 | 1 | 0.029 | 1 | 0.029 | |
| 0.01 | 0.99 | 0.065 | 0.971 | 0.056 | 0.99 | 0.047 | ||
| 0.02 | 0.981 | 0.09 | 0.895 | 0.11 | 0.981 | 0.078 | ||
| 0.1 | 0.867 | 0.172 | 0.848 | 0.2 | 0.886 | 0.162 | ||
| LBD | Original | 0.001 | 1 | 0.065 | 1 | 0.073 | 1 | 0.057 |
| 0.01 | 0.993 | 0.175 | 0.975 | 0.192 | 0.996 | 0.166 | ||
| 0.02 | 0.964 | 0.21 | 0.971 | 0.229 | 0.978 | 0.203 | ||
| 0.1 | 0.885 | 0.338 | 0.903 | 0.365 | 0.918 | 0.328 | ||
| Update | 0.001 | 0.946 | 0.04 | 0.957 | 0.039 | 0.967 | 0.033 | |
| 0.01 | 0.739 | 0.097 | 0.674 | 0.098 | 0.739 | 0.09 | ||
| 0.02 | 0.685 | 0.116 | 0.663 | 0.119 | 0.707 | 0.112 | ||
| 0.1 | 0.511 | 0.179 | 0.478 | 0.191 | 0.522 | 0.167 | ||
GLMnet = Generalized Linear Models with Convex Penalties; GBM = gradient boosting machine; Hybrid = Maximum prediction from either GLMnet or GBM; PPV = Positive predictive value
Note: We calculated bootstrapped standard errors for the GLMnet estimates. In all cases, the standard errors were substantially smaller (<0.005) than the estimates for sensitivity or PPV.
From: Results
NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.