G-computation of average treatment effects on the treated and the untreated

Table 1 Effect estimates obtained from g-computation using the illustrative example dataset^a (N = 7706)

	G-computation (via Monte Carlo Simulation)^b
	Point Estimate	Standard Error	95% Confidence Interval
Average Treatment Effect among the Treated (ATT)
Risk difference	−0.019	0.009	−0.040, −0.007
Odds ratio	0.773	0.112	0.607, 0.944
Average Treatment Effect among the Untreated (ATU)
Risk difference	−0.012	0.012	−0.036, 0.010
Odds ratio	0.910	0.133	0.678, 1.177
Average Treatment Effect (ATE)
Risk difference	−0.015	0.011	−0.036, 0.007
Odds ratio	0.884	0.127	0.676, 1.130

^aTreatment: education (1 = high school and beyond, 0 = less than high school); outcome: ever diagnosed with angina (1 = yes, 0 = no); covariates: age and gender
^bThe outcome model included all possible 2- and 3-way product terms between education and covariates. Standard errors and the 95% confidence limits were based on 500 bootstrap samples where the standard deviation of the 500 point estimates was taken as the standard error and the corresponding 2.5th and 97.5th percentiles were taken as the lower and upper limit of the 95% confidence interval

ISSN: 1471-2288