Table 2.
Applications of the l1-penalized least squares with hierarchical group LASSO variable selection (l1-PLS-HGL), l1-penalized least squares with group LASSO variable selection (l1-PLS-GL), adaptive contrast weighted learning (ACWL), and direct learning (D-learning) to data from male donors in the INTERVAL trial.
|
Assignment Percentages |
ITR Effects |
|||||
|---|---|---|---|---|---|---|
| Target Outcome | Method | 12 weeks | 10 weeks | 8 weeks | Donation | Deferral |
| Donation | l1-PLS-HGL | 0.1 (0.0) | 0.3 (0.0) | 99.6 (0.0) | 1.308 (0.004) | 0.026 (0.000) |
| l1-PLS-GL | 0.0 (0.0) | 0.3 (0.3) | 99.7 (0.3) | 1.311 (0.005) | 0.027 (0.000) | |
| ACWL | 0.0 (0.0) | 0.0 (0.0) | 100.0 (0.0) | 1.315 (0.000) | 0.027 (0.000) | |
| D-learning | 0.3 (0.1) | 0.3 (0.2) | 99.4 (0.2) | 1.307 (0.006) | 0.027 (0.000) | |
| Deferral | l1-PLS-HGL | 94.4 (0.6) | 5.5 (0.6) | 0.0 (0.0) | −1.188 (0.010) | −0.024 (0.000) |
| l1-PLS-GL | 99.7 (0.6) | 0.2 (0.5) | 0.0 (0.1) | −1.246 (0.006) | −0.024 (0.000) | |
| ACWL | 99.7 (0.6) | 0.3 (0.6) | 0.0 (0.0) | −1.244 (0.007) | −0.025 (0.000) | |
| D-learning | 95.7 (0.5) | 4.1 (0.5) | 0.2 (0.1) | −1.200 (0.011) | −0.024 (0.000) | |
Note: Means and standard deviations (in parenthesis) of assignment proportions in % and empirical ITR effects on donation and deferral outcomes across 100 repetitions of 5-fold cross-validation are reported. ITR effects measure the difference in the average outcome between donors whose assigned inter-donation intervals in the trial are optimal (with respect to the method used to estimate the ITR) and those whose assigned inter-donation intervals are non-optimal. A larger ITR effect on donation and a smaller ITR effect on deferral are more desirable. The first four and last four rows correspond to the target being maximizing total units of blood collected by the blood service, and minimizing the low Hb deferral rates, respectively.