A Method for Defining and Modeling Social Networks Inferred from Physical Distance and Related Information

Social ecologies are known to have major effects on adolescent behaviors of all kinds, including drug and alcohol use. In special cases, ecological properties can be measured using social network concepts—for example, if the individuals in question are all contained in some organization (a school, business, or club) or some small and easily accessible physical space, then the individuals within these contained units may be surveyed to determine who knows whom, friendship relationships, and so on. The question of how to obtain a map of social relationships in a physical neighborhood, however, has been little addressed, even though it is well-known that aspects of such units (with populations of as many as 4-5,000 individuals) are distinctive enough to have major effects on residents’ health and well-being generally, and on the course of development for adolescent residents more specifically.

In this study, we present a method for using physical proximity of residences and available auxiliary information (attending same school) to estimated probabilities of ties existing between any pair of the 250 adolescent participants in study of social and geospatial effects on adolescent problem behavior trajectories. The method uses results from Preciado, Snijders et al. (Social Networks 34, 18-31, 2012) to calculate tie probabilities in 1000 simulated networks, which are then analyzed via calculations of common local and global measures of structural characteristics (transitivity, behavioral clustering), as well as Exponential Random Graph Modeling (ERGM). Preliminary results suggest that this approach may offer a useful alternative to the measurement of social network ties in cases where methods like selecting names from a list are not feasible.