Simulating exponential random graph models for social networks: From local to global structure

Abstract

This paper reviews current progress in the development of exponential random graph models, a family of statistical models for modeling social networks. Full model specification requires formulation of a dependence hypothesis which proposes how possible network ties are conditionally dependent on each other. The dependence hypothesis in effect sets out the processes whereby network ties self organize into small micro-structures (network configurations). These can be construed as the building blocks of the global network structure. The paper presents three dependence assumptions that result in different types of graph distribution: (1) Bernoulli graph distributions whereby ties are assumed to ..