Influence Networks 11

calculations of centrality were used to test the hypotheses. The first, Centrality

1

, was calculated

with the indegree scores for communication ties within the influence network and standardized

by network size. In other words, Centrality

1

was calculated by dividing indegree communication

ties by the total possible communication ties scores within a network, resulting in a score of

‘normalized degree centrality’ (Borgatti, Everett, & Freeman, 1999).

The second, Centrality

2

, is a centrality measure used within the structural theory of social

influence (1998). The average indegree score of all network members, rather than network size,

is the basis for calculating Centrality

2

, which also includes indegree scores for each individual.

The relationship between an individual’s centrality score and Centrality

2

is that "an increase in

the density of interpersonal ties (mean indegree) lowers self-weight” (Friedkin, 1998, p. 96).

This calculation was proposed because it adjusts individual centrality so that the influencing

force of structurally central individuals is dampened in networks that contain many central

individuals, and the influencing force of structurally central individuals is enhanced in networks

where this position is rare. This test accommodates a contingent relationship between centrality

and influence, one that stipulates that the relationship is impacted by network properties. Adding

centrality at the second level enables a cross-level test between the level two variables and the

variables at level 1. Combining the first level equation with the second level produces a random

effect multilevel model (Bryk & Raudenbush, 1992; Hedeker et al., 1996; Snijders & Bosker,

2000)

i

:

Level 1

Behavioral Intent

*ij*

= ß

0* i*

+ ß

1* i*

Attitude* *

*ij*

+ ß

2* i*

Subjective Norm* *

*ij*

+ r

*ij*

Level 2

ß

0*i*

= ß

00

+ *u*

0*ij*

* *