Assessing the effectiveness of empirical calibration under different bias scenarios

Table 4 Average width of 95% confidence intervals with and without calibration for reference simulation (five negative controls, random Gaussian noise for measurement error)

	Ideal—Suitable negative controls		Random—Suitable negative controls		Random—Unsuitable negative controls
	Calibrated	Uncalibrated	Calibrated	Uncalibrated	Calibrated	Uncalibrated
Unmeasured Confounder	0.44	0.10	0.50	0.10	0.13	0.10
Model misspecification Quadratic term	0.61	0.11	0.62	0.11	0.13	0.11
Model misspecification Interaction term	0.39	0.11	0.44	0.11	0.13	0.11
Non-positivity	0.22	0.18	0.21	0.18	0.22	0.18
Measurement error	0.18	0.15	0.17	0.15	NA	NA

ISSN: 1471-2288