``Dynamic Stochastic Blockmodel Regression for Social Networks: Application to International Conflicts.''



Many social scientists theorize how various factors influence the dynamic process of network evolution. These theories explain the ways in which nodal and dyadic characteristics play a role in the formation and evolution of relational ties over time. We develop a dynamic model of social networks by combining a Hidden Markov model with a mixed-membership stochastic blockmodel that identifies latent groups underlying the network structure. Unlike existing models, we incorporate covariates that predict both the dynamic changes in the node membership of latent groups and the direct formation of edges between dyads. Our motivating application is the dynamic modeling of international conflicts. While most existing work assumes the decision to engage in militarized conflict is independent across states and static over time, we demonstrate that conflict patterns are driven by states' evolving membership in geopolitical coalitions. Changes in monadic covariates like democracy shift states between coalitions, generating heterogeneous effects on conflict over time and across states. The proposed methodology, which relies on a variational approximation to a collapsed posterior, is implemented through an open-source software package. (Last Revised, July 2018)


Olivella, Santiago, Tyler Pratt, and Kosuke Imai. ``NetMix: Mixed-membership Regression Stochastic Blockmodel for Networks.''

© Kosuke Imai
 Last modified: Sun Aug 19 10:37:47 EDT 2018