Missing data augmentation for bayesian exponential random multi-graph models

In this paper we present an estimation algorithm for Bayesian exponential random multi-graphs (BERmGMs) under missing network data. Social actors are often connected with more than one type of relation, thus forming a multiplex network. It is important to consider these multiplex structures simultaneously when analyzing a multiplex network. The importance of proper models of multiplex network structures is even more pronounced under the issue of missing network data. The proposed algorithm is able to estimate BERmGMs under missing data and can be used to obtain proper multiple imputations for multiplex network structures. It is an extension of Bayesian exponential random graphs (BERGMs) as implemented in the Bergm package in R. We demonstrate the algorithm on a well-known example, with and without artificially simulated missing data.