TABLE 1.
Summary of simulation parameters included in each simulation setting.
| Simulation setting | Total sample size | Covariate prevalence | α |
|
|
|
Stratification error rate | Errors across strata | Error discovery rate | Error discovery across treatment groups | Number of scenarios | |||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Effect of stratification errors on correlation between treatment groups (Section 2.1) | 1000 | 0.5 | 0 | 0 | 1, 3 | 0 | 0%, 5%, 10%, 15%, 20%, 25%, 30%, 35%, 40% | Equal | 100% | N/A | 18 | |||
| Estimating treatment effects when all stratification errors are discovered (Section 2.2) | 1000 | 0.5 | 0 | 0, 0.2 | 1, 3 | 0 | 1%, 20% | Equal, Unequal (three times higher in one stratum) | 100% | N/A | 16 | |||
| Reference case: Estimating treatment effects when there are no stratification errors (Section 2.2) | 1000 | 0.5 | 0 | 0, 0.2 | 1, 3 | 0 | 0% | N/A | N/A | N/A | 4 | |||
| Estimating treatment effects when only some stratification errors are discovered (Section 2.3) | 1000 | 0.5 | 0 | 0, 0.2 | 1, 3 | 0 | 1%, 20% | Equal, Unequal (three times higher in one stratum) | 50% | Equal, Unequal (three times higher in one group) | 32 | |||
| Estimating treatment by covariate interaction effects when all stratification errors are discovered (Section 2.4) | 1000 | 0.5 | 0 | 0, 0.1, 0.2 | 1 | 0.4 (when =0), 0.2 (when =0.1), 0 (when =0.2) | 1%, 20% | Equal, Unequal (three times higher in one stratum) | 100% | N/A | 12 | |||
| Estimating treatment by covariate interaction effects when only some stratification errors are discovered (Section 2.4) | 1000 | 0.5 | 0 | 0, 0.1, 0.2 | 1 | 0.4 (when =0), 0.2 (when =0.1), 0 (when =0.2) | 1%, 20% | Equal, Unequal (three times higher in one stratum) | 50% | Equal, Unequal (three times higher in one group) | 24 | |||
| Sensitivity simulations: Estimating treatment effects when only some stratification errors are discovered (Section 2.5) | 200, 1000 | 0.5 (when n = 200), 0.75 (when n = 1000) | 0 | 0 | 3 | 0 | 20% | Equal, Unequal (three times higher in one stratum a ) | 50% | Equal, Unequal (three times higher in one group) | 8 | |||
| Sensitivity simulations: Estimating treatment by covariate interaction effects when only some stratification errors are discovered (Section 2.5) | 200, 1000 | 0.5 (when n = 200), 0.75 (when n = 1000) | 0 | 0 | 1 | 0.4 | 20% | Equal, Unequal (three times higher in one stratum a ) | 50% | Equal, Unequal (three times higher in one group) | 8 |
a
When the covariate prevalence was 0.75, the error rate was three times higher in the more common stratum.