Modeling Concurrency and Selective Mixing in Heterosexual Partnership Networks with Applications to Sexually Transmitted Diseases

• Activity Number: 606
• Type: Contributed
• Date/Time: Wednesday, August 3, 2016 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Ryan Admiraal* and Mark Stephen Handcock
• Companies: Murdoch University and University of California at Los Angeles
• Keywords: heterosexual partnership networks ; concurrency ; selective mixing ; exponential-family random graph models ; constrained maximum likelihood estimation

Abstract

Stochastic models are instrumental in ascertaining the impacts of proposed interventions on STD transmission. Network-based models should reflect key characteristics in the population of interest, including concurrency and selective mixing. Current modeling approaches can only account for both concurrency and mixing by enforcing restrictions on either concurrency or mixing. We propose a new method that allows modelers to incorporate concurrency and mixing consistent with levels observed in the population of interest. Estimation of concurrency and mixing typically relies on cross-sectional data. In the context of heterosexual networks, we describe how this results in male- and female-specific reports that need not be consistent with each other. We provide a method to incorporate both sets of reports and jointly estimate degree distributions and mixing totals for males and females using constrained maximum likelihood methods. We demonstrate how to simulate heterosexual networks consistent with these degree distributions and mixing totals using exponential-family random graph models and apply to data from the National Longitudinal Study of Adolescent Health.