Influence Networks 19

participants who completed the survey (N=115). All reports of significance are 2-tailed. The

correlation between Centrality

2

and outdegree likelihood to comply was not statistically

significant (*r *= -.01, *p > *.05), but the correlation between Centrality

1

and outdegree likelihood

to comply was statistically significant (*r *= .26, *p < *.05). Although significant, the correlation

was positive, and the relationship between these two variables is not in the predicted direction.

Hence, the second hypothesis is not supported by these data. Table 1 contains the Pearson

Product-Moment correlations associated with these hypotheses.

--- Table 1 ---

The third hypothesis specifies a relationship between centrality and the regression weight

for attitude: as structural centrality increases, the weights associated with attitude will increase.

The measure of individual attitude was used for this test. This scale reflected good overall alpha

reliability (

Î±

= .84, N = 804) and also for each behavior, where alpha reliability ranged from .71

to .88, N = 115.

This hypothesis was tested by model comparison approach, the recommended procedure

within multilevel modeling (Bryk et al., 1992). The significance of the interaction is measured

by the degree to which the multilevel model fits the data. The â€˜model fitâ€™ is determined by

comparing the deviance score for a baseline model, in this case a model that contains the

elements at the individual level, with a model that contains a cross level interaction.

---Table 2---

For both types of centrality, the model containing the hypothesized relationship between

structural centrality and attitude weight did not result in a comparatively better fit of the data, as

evidenced by the failure of each model to significantly reduce the model deviance score in

comparison to the deviance score derived from the baseline model (Table 3 contains these