hergm: Hierarchical exponential-family random graph models

Abstract

We describe the R package hergm that implements hierarchical exponential-family random graph models with local dependence. Hierarchical exponential-family random graph models with local dependence tend to be superior to conventional exponential-family random graph models with global dependence in terms of goodness-of-fit. The advantage of hierarchical exponential-family random graph models is rooted in the local dependence induced by them. We discuss the notion of local dependence and the construction of models with local dependence along with model estimation, goodness-of-fit, and simulation. Simulation results and three applications are presented.