Table 2.

Mean Estimated Relative Excess Risk Due to Interaction and Empirical Coverage Rates of 95% Confidence Intervals Estimated From Generalized Estimating Equations^a Across Scenarios in a Simulation Study

Baseline Outcome Prevalence and RERI	Relative Risk			Fitted Model
	RR01b	RR10c	RR11d	Log Binomial GEE			Poisson GEE
				Mean $\widehat{\mathbf{RERI}}$	95% CI Coverage, %		Mean $\widehat{\mathbf{RERI}}$	95% CI Coverage, %
					δ	MOVER		δ	MOVER
0.01
5	2	3	9	5.25	94.7	94.1	5.25	94.7	94.1
10	2	3	14	10.60	95.4	93.7	10.60	95.4	93.8
0.1
1	1	1	2	1.00	95.1	95.5	1.00	95.1	95.5
1	2	3	5	0.98	95.5	95.5	0.98	95.5	95.6
3	1	2	5	3.00	95.5	95.4	3.00	95.5	95.4
0.2
1	1	1	2	1.00	95.5	95.5	1.00	95.5	95.5
1	1	2	3	0.99	94.9	95.1	0.99	95.0	95.0
2	2	2	4	1.00	94.9	95.0	1.00	95.0	95.1
0.4
1	1	1	2	1.00	95.2	95.1	1.00	95.1	95.2
0.5	1	1	1.5	0.50	95.3	95.2	0.50	95.2	95.2
0.6
0.5	1	1	1.5	0.50	95.2	95.5	0.50	95.2	95.4

Abbreviations: CI, confidence interval; GEE, generalized estimating equations; MOVER, method of variance estimates recovery; RERI, relative excess risk due to interaction; RR, relative risk.

^a GEE were estimated specifying an exchangeable working correlation, a log link, and either a binomial distribution or a Poisson distribution, with robust variance estimates. Results are based on 2,000 simulated data sets, each with 275 clusters.

^b RR₀₁, relative risk of the outcome in the group unexposed to X₁ and exposed to X₂ as compared with the doubly unexposed group.

^c RR₁₀, relative risk of the outcome in the group exposed to X₁ and unexposed to X₂ as compared with the doubly unexposed group.

^d RR₁₁, relative risk of the outcome in the doubly exposed group as compared with the doubly unexposed group.