Table 2.
Simulation results for Case 2A: the two-stage instrumental variable estimator
and the naïve rank estimator; all values have been multiplied by
|
Method | Parameter | Bias | ESE | ASE | ECR |
|---|---|---|---|---|---|---|
| 100 | Two-stage |
|
–2.5 | 24.5 | 25.4 | 94.6 |
|
0.2 | 30.4 | 29.8 | 95.4 | ||
| Naïve |
|
28.9 | 10.1 | 10.1 | 18.6 | |
|
–28.8 | 18.1 | 17.6 | 62.2 | ||
| 200 | Two-stage |
|
–2.5 | 17.7 | 17.4 | 91.7 |
|
0.6 | 20.8 | 20.3 | 93.3 | ||
| Naïve |
|
28.5 | 6.8 | 7.1 | 1.2 | |
|
–28.1 | 12.3 | 12.3 | 37.8 | ||
| 400 | Two-stage |
|
–0.3 | 11.7 | 12.3 | 95.8 |
|
–0.5 | 13.7 | 14.2 | 97.0 | ||
| Naïve |
|
27.9 | 4.7 | 5.0 | 0.0 | |
|
–27.7 | 8.8 | 8.6 | 13.0 |
Bias, average bias; ESE, empirical standard error; ASE, average of the estimated standard errors; ECR, empirical coverage rate of the 95% Wald confidence intervals.