Actor-Relation Effects

Effects that include the characteristics of the actors in the network are known as “actor-relation effects.” We use three basic effect types: sender, receiver, and homophily/heterophily (see Table 8.2 in Chapter 8). The sender and receiver parameters can be seen as main effects of attributes because they take into consideration only the attributes of one actor in the dyad with regard to a social tie. The homophily effect (used for binary and categorical attributes) is an interaction-type effect because the attributes of both actors in the dyad are modeled. The heterophily effect is for continuous attributes and is the converse of homophily (looking for difference rather than similarity).

Covariate Network Effects

Finally, one extra parameter included in the football network was a covariate network [Covariate Arc], which measures nominations of the best players in the team. Covariate networks are treated as exogenous in the model (see Chapter 8, Section 8.3).

Table 14.1. Two models for positive affect relations among schoolboys

estimates (SEs)


model A

model B

Purely structural effects (endogenous)


-2.67 (1.30)*

-2.91 (1.36)*


2.19 (0.29)*

2.30 (0.28)*

Simple 2-path

-0.13 (0.05)*

-0.11 (0.04)*

Popularity spread [AinS]a

-0.68 (0.29)*

-0.78 (0.29)*

Activity spread [AoutS]

-0.71 (0.30)*

-0.76 (0.29)*

Path closure [AT-T]

1.57 (0.13)*

1.54 (0.12)*

Multiple 2-paths [A2P-TD]

0.08 (0.08)

0.06 (0.07)

Actor-relation effects (exogenous)

Heterophily® - personal attitudes

-0.15 (0.06)*

-0.25 (0.08)*

Heterophily - perceived attitudes

0.16 (0.09)

Sender - personal attitudes

0.01 (0.08)

-0.26 (0.12)*

Sender - perceived attitudes

0.35 (0.12)*

Receiver - personal attitudes

0.03 (0.08)

0.33 (0.15)*

Receiver - perceived attitudes

-0.37 (0.15)*

Homophily - ethnoculture

0.48 (0.14)*

0.46 (0.12)*

Sender - ethnoculture

-0.06 (0.15)

-0.04 (0.15)

Receiver - ethnoculture

-0.32 (0.16)

-0.35 (0.17)*

Heterophily - socioeconomic status (SES)

-0.16 (0.10)

-0.17 (0.10)

Sender - SES

-0.29 (0.13)*

-0.27 (0.13)*

Receiver - SES

0.36 (0.14)*

0.37 (0.15)*

* Significant effect (i.e., parameter estimate is greater than two times the standard error in absolute value, see Section 12.5.1 for details)

a Even if the popularity spread [A-in-S] and activity spread [A-out-S] parameters are not significant, it is still useful to include them in the model because they control for popularity and activity. If these parameters are included and there are still substantial sender and receiver effects, then this is stronger support for the impact of individual-level attributes as explaining ties that are sent or received. b Remember that the homophily parameters presented in Chapter 8 have different valence for binary and continuous variables (also noted in the example in Chapter 13). For the binary variable, we examine pairs of nodes to determine whether they both have the presence of “1” as their attribute, so homophily is represented by a significant positive parameter, and we refer to this here as homophily. For continuous variables, we examine the absolute difference between the continuous attributes for the dyad, and this difference effect is called “heterophily” - that is, a significantly positive parameter indicates that actors are different in an attribute, whereas a significant negative parameter estimate indicates that actors are similar people and have less of a difference in the attribute.

< Prev   CONTENTS   Source   Next >