A Hidden Markov Model for Animal Soical Networks

• Activity Number: 541
• Type: Contributed
• Date/Time: Wednesday, August 3, 2016 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Meridith L. Bartley* and Ephraim M. Hanks and David P. Hughes
• Companies: Penn State University and Penn State University and Penn State University
• Keywords: Ecological Modelling ; Contact Network ; HMM ; Bayesian Estimatation ; Ant Trophallaxis ; MCMC

Abstract

Contacts between animals drive critical ecological processes such as gene flow and the spread of infectious disease. Current models for animal contact networks consider data observed at discrete times and often assume that the properties of the contact network (e.g., the rate at which contacts occur) are homogeneous in time. We propose a novel model for animal interactions driven by a latent Hidden Markov Model (HMM) . This HMM structure allows us to model varying rates of interactions that are dictated by underlying biological behaviors within (or between) species. We develop a Markov chain Monte Carlo algorithm to perform Bayesian inference based on continuous-time observations of the social network. We illustrate our approach through the analysis of trophallaxis interactions between carpenter ants (Camponotus pennsylvanicus) over eight days of continuous time observations.