Example 2: Football Team
We examine two models for aggression relations within an all-male football team. Remember, in this network, the sender of a tie is the aggressor, and the receiver of the tie is the victim. Results are presented in Table 14.2.
Results: Purely Structural Effects. In contrast to the schoolboys, the parameter estimates for reciprocity and closure are not significant for the football team. This is not surprising for a network based on negative ties (i.e., aggression). Instead there is both a significant and positive effect for activity and popularity spread, indicating that some players are overly “active” as aggressors, and some are overly “active” as victims of aggression.
Results: Actor-Relation Effects. The control variables offer some insights for the footballers. In model B, a negative and significant receiver effect for relationship indicates that single men were substantially more likely to be the recipient of aggression. Furthermore, aggression relations were much more likely between players of differing levels of experience, as noted by the significant positive heterophily effect for experience. Interestingly, it is the inclusion of perceived attitudes that brings out the heterophily effect for experience (compare models A and B).
For this football team, we see an effect of perceived attitudes that is independent of personal attitudes. Again, the focus is on the difference made by including perceived attitudes. Unlike the schoolboys’ network, for the footballers there are no significant effects for personal MAI for either model. Furthermore, although the parameter values of personal MAI change when including perceived MAI attitudes, the personal attitudes have no substantial impact on social tie formation. However, perceived MAI attitudes are significantly related to both sending and receiving ties. Therefore, footballers with high perceived MAI are significantly more likely to be aggressive toward others. Players with high perceived MAI are significantly more likely to be victims of aggression. Interestingly, there is no homophily effect for perceived MAI. Here, we have the extreme situation where perceptions about the collective determine behaviors independently of personal motivations.
Results: Covariate Network Effects. Finally, there is no effect of best player covariate network on the aggression network.