Table 2.

Comparing the total MC variances based on M = 500 new importance samplings between the bootstrap approach and the exact approach. Results for α and φ are not presented as the estimates are constant. A single 30-day epidemic in 100 households was simulated, with b = 0.01, p = 0.05, α = 0.2, φ = 0.7, θ = 0.5, and OR=0.6. θ is assumed known in estimation. The burn-in number of importance samples is K₀ = 200.

K	Traditional MCEM (J = 2)						Hybrid EM-MCEM (J = 100)
	b(×10−11)		p(×10−9)		OR(×10−8)		b(×10 −11)		p(×10−9)		OR(×10−8)
	Boot.	Exact	Boot.	Exact	Boot.	Exact	Boot.	Exact	Boot.	Exact	Boot.	Exact
500	9.77	9.64	17.88	18.20	13.52	13.80	3.13	3.37	6.19	6.20	5.34	5.43
1000	4.91	5.14	8.64	9.29	6.32	6.53	1.44	1.59	2.78	2.94	2.54	2.67
2000	2.40	2.29	4.44	4.70	3.46	3.46	0.74	0.87	1.48	1.65	1.41	1.50
5000	0.91	0.86	1.55	1.51	1.21	1.15	0.31	0.31	0.60	0.56	0.56	0.49