Overlapping Activity Patterns and Community Detection in Ecological Networks

• Activity Number: 245 - Bayesian Inference in the Life Sciences and Medicine
• Type: Contributed
• Date/Time: Monday, July 29, 2019 : 2:00 PM to 3:50 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Wenna Xi* and Catherine A. Calder and Christopher R. Browning
• Companies: The Ohio State University and The Ohio State University and The Ohio State University
• Keywords: Bayesian modeling; latent Dirichlet allocation; network analysis; social science

Abstract

The extent to which activity spaces -- the collection of an individual’s routine activity locations -- overlap provides important information about the function of a city and its neighborhoods. To study the patterns of overlapping activity spaces and to detect communities of individuals based on their shared locations, we introduce the notion of an ecological network, a type of two-mode network with the two modes being individuals and routine activity locations. Specifically, we identify latent activity pattern profiles, which, for each community, summarize its members’ probability distribution of going to each location, and community assignment vectors, which, for each individual, summarize his/her probability distribution of belonging to each community. Using data from the Adolescent Health and Development in Context (AHDC) Study, we employ latent Dirichlet allocation (LDA) to identify activity pattern profiles and community assignment vectors. We then explore differences across neighborhoods of Columbus, OH in the strength, and within-neighborhood consistency of community assignment, paying particular attention to the association between race and these measures.