. Author manuscript; available in PMC: 2009 May 11.

Published in final edited form as: Ann Appl Stat. 2007 Jun 1;1(1):211–228. doi: 10.1214/07-AOAS105

Table 3.

Comparison of type I error and power between the permutation test and the asymptotic test for models with only b and p₁. The community is composed of 100 households of size 5. Results are based on 2000 simulations. 2000 permuted samples were drawn for each test

b	CPI	p₁^a	SAR₁	N_idx^b	N_tot^c	Asymptotic	Simple permutation	Refined permutation
0.0005	0.015	0.0	0.0	7	7	0.037	0.042	0.046
		0.010	0.039	7	8	0.51	0.52	0.53
		0.020	0.078	7	9	0.78	0.77	0.78
		0.030	0.11	7	10	0.87	0.86	0.87
0.0010	0.03	0.0	0.0	13	14	0.031	0.047	0.047
		0.010	0.039	13	16	0.59	0.64	0.64
		0.015	0.059	13	17	0.78	0.81	0.81
		0.020	0.078	13	18	0.88	0.90	0.90
0.0050	0.14	0.0	0.0	51	66	0.037	0.049	0.053
		0.005	0.020	51	69	0.43	0.45	0.47
		0.010	0.039	51	74	0.85	0.85	0.86
		0.015	0.059	51	78	0.97	0.97	0.97

Type I errors are reported when p₁ = 0.

N_idx is the average number of index cases.

N_tot is the average total number of cases.