Node, Relationship, and Network Attributes

Node Attributes

Most network studies conducted by social scientists include data describing the nodes, often termed actors, connected by the relationships that make up a network. Nodal attribute data can encompass virtually any individual-level measurement; the subject matter of any given study guides the selection of attributes. Attributes can include information deemed relevant to predicting network structure or formation, including background characteristics such as an individual’s age, gender, education, or race/ethnic identity, and indicators of membership or official position within an organization or association. Other actor attributes regarded as responses to network location—for example, as outcomes of social influence processes or social support provision—may also be included. Examples are behaviors such as smoking, delinquent activity, or alcohol use; achievements like academic performance or awards; and attitudes or sentiments such as happiness, life satisfaction, political orientation, or tastes for musical genres.

Visualizations of networks can use attribute values to distinguish nodes by size, color, shading, or shape. Exploration of patterns in visual renderings may suggest differences between types of actors situated in central and peripheral locations within a network, or clusters of actors that tend to share similar attribute values. Statistical analyses draw on nodal attributes when identifying types of actors that are more or less apt to be involved in relationships, such as factors predicting student popularity in school classrooms.

Many network composition measures used in egocentric network studies are based on the average attribute values for the set of “alter” nodes that lie within a focal “ego” node’s network; examples include the proportion of liberals among a node’s alters or their average age. Some measures of network diversity are based, correspondingly, on variability in attribute values for those alters in an actor’s egocentric network.

[Page 1064]One-mode cross-sectional network studies examine relationships linking nodes of one type—like students or countries—on a single occasion. These would include one set of attribute values describing each node. Two-mode designs involving relationships between two types of entities, such as memberships of students in extracurricular groups, may introduce separate sets of attributes for the nodes of different types. Extracurricular groups might be distinguished by their size, gender composition, and principal activity (athletic, arts-related, service-oriented), for instance, and students by their background characteristics.

Longitudinal network studies measure relationships among the nodes in a network repeatedly at points in time. Such studies can include both time-constant attributes like gender and time-varying ones such as smoking behavior at each occasion of measurement. Longitudinal designs are better suited than cross-sectional ones to adjudicating between alternative accounts for associations between nodal attributes and relationships. Such associations might reflect influences of the alters related to an actor, for example, characteristics of one’s close friends that prompt smoking cessation. They can, of course, also reflect selection processes—for instance, when an actor breaks off existing relationships and initiates new ones because she or he wants to stop smoking.

...