Table 12.

Homophily bias regression: beta coefficients from closeness and in-degree simulations.

Variables	Model 1	Model 2	Model 3	Model 4
	In-degree	Out-degreea	High behavioral	Low behavioral
Intercept	2.276*	1.516**	.111*	.234***
	(.92)	(.53)	(.05)	(.02)
Correlation with centrality	.87*	.628*	.001	.08***
	(.43)	(.29)	(.01)	(.02)
Correlation type (0 = closeness1 = in-degree)	.168***	.107***	5e−04	.005**
	(.04)	(.03)	(.001)	(.002)
Correlation with Centrality*Correlation type	.391***	.219***	.005***	.009**
	(.07)	(.06)	(.001)	(.003)
Directed	.151	−.074	−.02	−.033***
	(.35)	(.23)	(.02)	(.01)
Correlation with Centrality*Directed	−.001	−.118	−.002	−.029***
	(.16)	(.13)	(.004)	(.01)
Log of Size	−.326	−.253*	−.013	−.029***
	(.19)	(.11)	(.01)	(.004)
Correlation with Centrality*Log of Size	−.145	−.112	4e−04	−.009*
	(.09)	(.06)	(.002)	(.005)
In-degree Std. Dev.	−.03	.033	−.001	−.002
	(.07)	(.05)	(.004)	(.001)
Correlation with Centrality*In-degree Std. Dev.	−.001	.022	−.001	−.001
	(.03)	(.03)	(.001)	(.002)
N	96	88	96	96
Networks	12	11	12	12

Note: The regression uses the beta slopes from each line as the dependent variable. The direction of the bias is ignored when calculating the regressions. The betas represent the expected increase in bias for a 10% increase in the amount of missing data. Larger numbers mean larger bias with more missing data. The correlation with centrality takes four values: −.75, −.25, .25, and .75.

^aRC Elite network removed from regression as it is an extreme outlier.