Table 1.

Simulation scenarios with homogeneous/heterogeneous clusters.

Location		Homogeneous clusters									Heterogeneous clusters
		Rural			Mixed			Urban			Two clusters		Three clusters
Cases	Size	POP	RR	P	POP	RR	P	POP	RR	P	POP	H	POP	H
600	1	2675	192.89	1.998	710196	2.85	1.946	786178	2.73	1.941	788853	190.16	1499049	190.16
	2	22911	27.03	1.992	817050	2.70	1.943	1072181	2.43	1.932	1095092	24.6	1912142	24.6
	4	132343	7.05	1.979	1108440	2.40	1.931	2953077	1.81	1.881	3085420	5.24	4193860	5.24
	8	204829	5.35	1.971	1352284	2.24	1.923	5018909	1.63	1.836	5223738	3.72	6576022	3.72
	16	360275	3.9	1.961	1684327	2.1	1.914	7627173	1.53	1.785	7987448	2.37	9671775	2.37
6000	1	2675	23.73	20.27	710196	1.45	20.09	786178	1.43	20.08	788853	22.3	1499049	22.3
	2	22911	4.96	20.25	817050	1.42	20.09	1072181	1.36	20.05	1095092	3.6	1912142	3.6
	4	132343	2.21	20.21	1108440	1.36	20.04	2953077	1.22	19.88	3085420	0.99	4193860	0.99
	8	204829	1.92	20.18	1352284	1.32	20.02	5018909	1.17	19.73	5223738	0.75	6576022	0.75
	16	360275	1.66	20.15	1684327	1.29	19.99	7627173	1.15	19.57	7987448	0.51	9671775	0.51

Note: POP is the total population in the clusters. RR is the relative risk of the clusters. P is the incidence (×10⁻⁵) out of the clusters. Because the case numbers follow a Poisson distribution under the null hypothesis, the incidence rate also reflects random fluctuation out of the clusters. H is the difference between the maximal RR and minimal RR that reflects the strength of the heterogeneity among the clusters. Heterogeneity among the clusters becomes lower as the cluster number, total case number and cluster size grow.