Table 5.:
Performance as measured by AUC and negative log-likelihood for the three Super Learners with the following libraries: machine learning algorithms with only baseline covariates, augmenting this library with hdPS, and only the machine learning algorithms but with both baseline and hdPS screened covariates. (See Table 1).
| Data set | Performance Metric | Super Learner 1 | Super Learner2 | Super Learner 3 |
|---|---|---|---|---|
| NOAC | AUC | 0.7652 | 0.8203 | 0.8304 |
| NSAID | 0.6651 | 0.6967 | 0.6975 | |
| VYTORIN | 0.6931 | 0.6970 | 0.698 | |
| NOAC | Negative Log-likelihood | 0.5251 | 0.4808 | 0.4641 |
| NSAID | 0.6099 | 0.5939 | 0.5924 | |
| VYTORIN | 0.4191 | 0.4180 | 0.4171 |