A Bipartite Dynamic Networks Approach to Place-Based Risk Among Urban Adolescents

Exposure to peers who use alcohol and drugs is a known risk factor for subsequent use by adolescents. However, the context of this exposure is also important, including, for example, who else may be present, the sorts of activities that occur, substance availability, and individual variability in characterization of a location’s riskiness. Using a bipartite graph framework, where the two classes of nodes are individual adolescents and urban locations, we present a mathematical framework that describes the interrelations over time between the substance use of individual adolescents and the characteristics of urban locations they spend time in.  Because  it is quantitative, this approach is amenable to theoretical analysis via simulation. Furthermore, suitable data may be used to estimate the model interrelationships quantitatively, with a variant of Snijders’ Stochastic Actor-Based (SAB) modeling as implemented in the R software package RSiena. We provide results from a series of theoretical simulations based on our previous work, predicting escalation of individual drug and alcohol individual adolescents’ outcomes on the basis of characteristics of their personal (ego-centered) networks and pattern of use of more and less risky urban locations, as well as characteristics of both the individuals and the locations they occupy. Preliminary SAB statistical models will also be presented, using early data from the recently-initiated research project the forms the core of this symposium.