Table 5.

Clustering Results for 5 Largest Add Health Networks by Sample Size

			Bias a				SE b
Net ID	Statistic	True Value	10% Sample	25% Sample	50% Sample	75% Sample	10% Sample	25% Sample	50% Sample	75% Sample
16	Modularity	.575	−.038	−.041	−.052	−.049	.022	.023	.017	.016
17	Modularity	.667	−.009	−.014	−.014	−.013	.019	.014	.013	.013
18	Modularity	.611	−.045	−.051	−.052	−.051	.026	.021	.016	.015
19	Modularity	.545	−.034	−.028	−.035	−.037	.016	.016	.013	.013
20	Modularity	.627	−.020	−.025	−.029	−.026	.026	.017	.012	.012
16	Transitivity	.140	.027	.019	.015	.015	.015	.014	.011	.011
17	Transitivity	.161	.028	.024	.021	.022	.019	.016	.015	.015
18	Transitivity	.178	.025	.013	.013	.015	.024	.019	.014	.013
19	Transitivity	.141	.012	.007	.006	.008	.017	.013	.011	.011
20	Transitivity	.138	.020	.010	.004	.006	.025	.017	.015	.014
16	Proportion 102 Triad	.015	.000	.000	.000	.000	.001	.000	.000	.000
17	Proportion 102 Triad	.010	.000	.000	.000	.000	.000	.000	.000	.000
18	Proportion 102 Triad	.011	.000	.000	.000	.000	.001	.000	.000	.000
19	Proportion 102 Triad	.012	.000	.000	.000	.000	.000	.000	.000	.000
20	Proportion 102 Triad	.006	.000	.000	.000	.000	.000	.000	.000	.000
16	Proportion 300 Triadc	.424	.078	.043	.042	.036	.058	.043	.037	.033
17	Proportion 300 Triadc	.219	.041	.032	.026	.026	.034	.026	.022	.020
18	Proportion 300 Triadc	.309	.037	.019	.024	.023	.047	.031	.025	.023
19	Proportion 300 Triadc	.267	.022	.012	.009	.012	.041	.028	.021	.020
20	Proportion 300 Triadc	.070	.011	.005	.002	.002	.015	.010	.008	.007

Note: The values for Bias and SE are calculated over 30 independent samples, where each sample yields one estimate of the network measure. All estimates come from the ENC model.

^aBias=E(estimates)−True Value.

^bThe Standard Error is the standard deviation of the sampling distribution.

^cValues are 10⁻⁵