Statistical Frontiers in Dynamic Network Modeling

This project explores statistical advancements in Relational Event Modeling (REM) to analyse dynamic networks where interactions occur in continuous time. REMs can be used to study anything from social interactions, financial transactions, patient transfers and alien species invasions, capturing fine-grained temporal dependencies.

This project will focus on making REMs more interpretable, faster and more robust. In particular, it will focus on three challenges: (i) The role of time – understanding how network effects evolve across different time scales and how past events influence future interactions, in connection with novel goodness-of-fit strategies; (ii) Computational bottlenecks – improving efficiency in estimating network covariates to enable scalable analysis of large datasets; (iii) Causal inference in dynamic networks – such as disentangling homophily (preference for similar others) from social influence (behavioral adaptation due to interactions).

Building on prior work in theory and applications, this research will integrate modern statistical learning techniques to enhance REM's capabilities. The goal is to develop more robust causal inference methodologies, improve computational efficiency, and refine the modelling of temporal dependencies in dynamic networks. These advancements will contribute to the broader field of statistical network science, making REM a more powerful tool for understanding complex event-driven systems.