Table 2 Estimates of performance measures for confounder identification approaches using plasmode simulation, values in parentheses are Monte Carlo standard errors when covariates are adjusted with regression
Performance measure | Outcome cause | Common cause | All pre-treatment |
|---|---|---|---|
Average treatment effect estimate | 0.020 | 0.021 | 0.021 |
Bias | -0.780 (0.0094) | -0.791 (0.0094) | -0.790 (0.0097) |
Coverage | 0.040 (0.0088) | 0.040 (0.0088) | 0.050 (0.0097) |
Bias eliminated coverage | 0.960 (0.0088) | 0.958 (0.0090) | 0.954 (0.0094) |
Empirical standard error | 0.211 (0.0067) | 0.211 (0.0067) | 0.217 (0.0069) |
MSE | 0.652(0.0150) | 0.653 (0.0150) | 0.654 (0.0155) |
Model based standard error | 0.209(0.0003) | 0.208 (0.0003) | 0.213 (0.0003) |
Relative % error in standard error | -0.847 (3.1422) | -1.399 (3.1246) | -1.72 (3.1147) |
Power of 5% test | 0.042 (0.0090) | 0.046 (0.0094) | 0.052 (0.0099) |