Table 3:
Testing Values (Standard Errors) on the Mixture of Subgroups (ncalib = 50)
| Testing Subgroup 1 Probability |
|||||
|---|---|---|---|---|---|
| type | 0.1 | 0.25 | 0.5 | 0.75 | 0.9 |
| LB-ITR | 1.665 (0.0067) | 1.537 (0.00618) | 1.444 (0.00412) | 1.545 (0.00537) | 1.679 (0.00585) |
| ℓ1-PLS | 1.182 (0.00191) | 1.264 (0.0014) | 1.399 (0.000591) | 1.537 (0.000333) | 1.624 (0.000781) |
| Standard ITR | 1.143 (0.00434) | 1.232 (0.00329) | 1.383 (0.0015) | 1.535 (0.000543) | 1.632 (0.00142) |
| RCT-DR-ITR | 1.267 (0.0066) | 1.305 (0.00423) | 1.395 (0.00256) | 1.52 (0.00212) | 1.614 (0.00234) |
| CTE-DR-ITR | 1.16 (0.00409) | 1.247 (0.00323) | 1.388 (0.00137) | 1.534 (0.00055) | 1.628 (0.00149) |
1
Testing subgroup 1 probability = 0.75 is the same as the training one.
2
Values (larger the better) can be comparable for the same subgroup 1 probability but incomparable across different subgroup 1 probabilities
3
LB-ITR maximizes the testing value function over the linear ITR class. The corresponding testing value is the best achievable among the linear ITR class.