2020 Sunbelt Workshop: Multilevel and Hierarchical Exponential-Family Random Random Graph Models with Local Dependence


In social networks, ties depend on other ties owing to social processes that give rise to transitive closure and other forms of closure. While in small social networks ties can depend on all other ties, in large social networks ties do not depend on all other ties but depend on a subset of other ties, because social networks are more local than global in nature. A simple class of models that respects the local nature of social networks assumes that actors are divided into subsets and ties depend on other ties within the same subset, but do not depend on ties outside of the subset. If the subsets are observed, the network is a special case of a multilevel network. Otherwise, the subsets have to be learned from observed social networks.

The proposed workshop focuses on ERGMs with local dependence:

  1. ERGMs with observed subsets of actors (multilevel ERGMs), implemented in R package mlergm.
  2. ERGMs with unobserved subsets of actors (hierarchical ERGMs), implemented in R package hergm. The basic ideas of ERGMs with local dependence will be introduced along with examples, and workshop attendees will be given the opportunity to use R packages mlergm and hergm with worked-out examples.