Table A3.

Maximum percent missing to remain under target bias of .25: Alternative measures of bicomponent size and ConCorr.

Network	Correlation with centrality	Bicomponent size (divided by size)	Bicomponent size (divided by component size)	ConCorr: depth set to 3	ConCorr: depths allowed to vary
Interlock	−.75	44	60	^a	^a
Interlock	.75	10	12	42	36
Prison	−.75	31	41	^a	^a
Prison	.75	16	15	60	^a
Sorority	−.75	34	45	^a	^a
Sorority	.75	17	19	60	54
6th graders	−.75	^a	^a	^a	^a
6th graders	.75	67	64	^a	^a
Coauthor	−.75	34	57	^a	^a
Coauthor	.75	10	11	44	32
Prosper	−.75	43	53	^a	^a
Prosper	.75	32	32	^a	^a
Co-citation	−.75	^a	^a	^a	^a
Co-citation	.75	23	20	53	38
Elites	−.75	36	53	^a	^a
Elites	.75	12	4	27	10
HS 13	−.75	69	^a	^a	^a
HS 13	.75	52	50	^a	^a
BioTech	−.75	38	^a	^a	^a
BioTech	.75	14	10	32	22
HS 24	−.75	66	^a	^a	^a
HS 24	.75	49	49	^a	68
CSprings	−.75	31	54	^a	^a
CSprings	.75	11	8	35	30
Mean (Std Dev)	−.75	47.2 (16.5)	59.4 (10.6)	70 (0)	70 (0)
Mean (Std Dev)	.75	26.1 (19.5)	24.5 (19.7)	52.8 (16.2)	47.5 (22)

Cases where percent missing is above 70, our observed maximum. In these cases, 70 is used to calculate overall means. The maximum percent missing was calculated based on a quadratic fit to the data.