Skip to main content
. 2017 Oct 26;104(4):881–899. doi: 10.1093/biomet/asx056

Table 3.

Simulation results for Case 2B: the instrumental variable estimators of Cases 2B and 1A and the naïve rank estimator; all values have been multiplied by Inline graphic

      Inline graphic Inline graphic
Inline graphic Method Parameter Bias ESE ASE ECR Bias ESE ASE ECR
800 Case 2B IV Inline graphic –3.5 25.7 25.2 96.2 –1.3 12.4 12.1 94.8
    Inline graphic 0.2 5.4 5.5 96.2 –0.2 4.8 4.9 95.2
  Case 1A IV Inline graphic –15.2 26.4 26.1 92.0 –2.4 14.3 13.8 92.6
    Inline graphic 0.3 5.6 5.6 96.6 –0.1 5.1 5.3 95.8
  Naïve Inline graphic 28.8 5.7 5.9 0.2 23.4 5.3 5.3 0.6
    Inline graphic –2.7 4.4 4.6 90.6 –2.3 4.4 4.6 92.6
1600 Case 2B IV Inline graphic –1.0 17.7 17.3 94.9 –0.4 8.7 8.5 93.6
    Inline graphic 0.1 3.8 3.8 94.7 0.0 3.4 3.5 95.9
  Case 1A IV Inline graphic –13.4 18.4 18.2 89.6 –1.4 9.8 9.7 94.2
    Inline graphic 0.2 3.8 3.9 95.2 0.0 3.6 3.7 96.4
  Naïve Inline graphic 29.2 3.9 4.2 0.0 23.3 3.7 3.7 0.0
    Inline graphic –2.8 3.2 3.2 85.2 –2.1 3.1 3.2 90.2

Bias, average bias; ESE, empirical standard error; ASE, average of the estimated standard errors; ECR, empirical coverage rate of the 95% Wald confidence intervals; Case 2B IV, instrumental variable method of Case 2B; Case 1A IV, instrumental variable method of Case 1A.