Table 1.
Simulation study results with n = 300; Maximum ASB, Standardized Bias, Standardized Var, and Standardized MSE all multiplied by 100, tm(Y, X) indicates the MSE-based method that uses outcome information, tm(X) indicates the MSE-based method that does not use outcome information, ta(0.20) and ta(0.10) indicate the ASB-based methods that use a 0.20 and 0.10 threshold, respectively, tp(0.05) and tp(0.10) indicate the p-value based methods that use a 0.05 and 0.10 threshold, respectively
| n=300 | ||||
|---|---|---|---|---|
| Setting (i) | Setting (ii) | Setting (i) | Setting (ii) | |
| Method | # of Trees | Maximum ASB | ||
| Unweighted | – | – | 70.02 | 70.01 | 
| TWANG | 502 | 494 | 23.49 | 23.38 | 
| tm(Y, X) | 176 | 446 | 40.55 | 23.44 | 
| tm(X) | 452 | 445 | 23.62 | 23.52 | 
| ta(0.20) | 461 | 453 | 23.97 | 23.86 | 
| ta(0.10) | 501 | 493 | 23.50 | 23.39 | 
| tp(0.05) | 378 | 369 | 25.38 | 25.35 | 
| tp(0.10) | 445 | 436 | 24.17 | 24.08 | 
| Standardized Bias | Standardized Var | |||
| Unweighted | 11.56 | 36.49 | 1.36 | 1.46 | 
| TWANG | 4.27 | 12.36 | 1.61 | 1.62 | 
| tm(Y, X) | 5.78 | 12.39 | 1.60 | 1.60 | 
| tm(X) | 4.31 | 12.46 | 1.60 | 1.60 | 
| ta(0.20) | 4.37 | 12.62 | 1.60 | 1.60 | 
| ta(0.10) | 4.27 | 12.36 | 1.61 | 1.62 | 
| tp(0.05) | 4.62 | 13.42 | 1.56 | 1.56 | 
| tp(0.10) | 4.40 | 12.74 | 1.59 | 1.59 | 
| Standardized MSE | Design Effect | |||
| Unweighted | 2.69 | 14.77 | 1 | 1 | 
| TWANG | 1.79 | 3.14 | 1.22 | 1.21 | 
| tm(Y, X) | 1.93 | 3.13 | 1.10 | 1.21 | 
| tm(X) | 1.78 | 3.15 | 1.21 | 1.21 | 
| ta(0.20) | 1.79 | 3.19 | 1.21 | 1.21 | 
| ta(0.10) | 1.79 | 3.14 | 1.22 | 1.21 | 
| tp(0.05) | 1.78 | 3.36 | 1.19 | 1.18 | 
| tp(0.10) | 1.79 | 3.21 | 1.20 | 1.20 |