Statistical Models for Dynamic Networks with Endogenous Vertex Migration

University Of California-Irvine, Irvine CA

Investigators

Abstract

This research project will develop models of complex systems in which the movement of social entities themselves (either into or out of a system of interest or among subsystems) is endogenously related to the relationships among those entities. Endogenous migration is critical to understanding important social phenomena ranging from recruitment into and turnover in organizations to the mass convergence of volunteer responders that occurs when disaster strikes. Endogenous migration is an important driver of the heterogeneity that can challenge conventional models of social network structure. This project will address current limitations in the modeling of networks with endogenous migration. The development of these models will advance modeling of complex social systems. Although the primary impact of this research will be within the statistical and social science communities, the tools and techniques to be developed also will be applicable to problems in biology, computer science, and engineering. The resulting insights will have direct policy relevance for groups or organizations dealing with important societal issues, such as emergency responses. The project also will make contributions via student education and training, the creation of instructional materials, and freely available software tools for use by government, industry, researchers, and the general public. This project will develop new families of statistical models for studying social networks with endogenous migration processes. The project will build on the well-known exponential family random graph model and related network modeling frameworks to integrate migration processes. The investigator will develop of new classes of models for dynamic relational data with endogenous migration and for cross-sectional data arising from unobserved migration-dependent processes. The investigator will evaluate and test these new model classes using social network data. A variety of data sets will be used as testbeds for the new models, including social media data, disaster-related data, and water and polymer data. Testbed applications will be used to evaluate the models and also facilitate the communication of results across disciplines. Broad access to these models will be ensured by the development of freely available software toolkits and the creation of training materials and workshops. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.