TABLE A3.
Single Confounder Removal: simulation results for B(−Xj) using nonparametric estimation with n = 2000;
Variable | 1 | 2 | 3 | 4 | 5 |
---|---|---|---|---|---|
Setting 1 | |||||
Truth | 0.347 | 0.225 | 0.187 | 0.196 | 0.044 |
Estimate | 0.401 | 0.215 | 0.189 | 0.174 | 0.022 |
Bias | −0.054 | 0.011 | −0.001 | 0.022 | 0.022 |
ESE | 0.044 | 0.045 | 0.035 | 0.032 | 0.016 |
Setting 2 | |||||
Truth | 0.580 | 0.026 | 0.305 | 0.030 | 0.059 |
Estimate | 0.608 | 0.037 | 0.293 | 0.042 | 0.020 |
Bias | −0.029 | −0.010 | 0.011 | −0.012 | 0.040 |
ESE | 0.052 | 0.018 | 0.048 | 0.021 | 0.016 |
Setting 3 | |||||
Truth | 0.389 | 0.219 | 0.031 | 0.296 | 0.065 |
Estimate | 0.436 | 0.184 | 0.072 | 0.286 | 0.022 |
Bias | −0.046 | 0.035 | −0.041 | 0.010 | 0.042 |
ESE | 0.050 | 0.036 | 0.035 | 0.047 | 0.018 |
Setting 4 | |||||
Truth | 0.298 | 0.211 | 0.198 | 0.195 | 0.098 |
Estimate | 0.400 | 0.216 | 0.189 | 0.173 | 0.022 |
Bias | −0.102 | −0.005 | 0.009 | 0.021 | 0.076 |
ESE | 0.046 | 0.047 | 0.037 | 0.032 | 0.017 |
ESE = empirical standard error (standard deviation of estimates across the 1,000 replications)