Statistical network analysis of time-ordered social interaction events with internal structure

• Applicant: Dr. Jürgen Lerner
• Subject Area: Statistics and Econometrics ; Empirical Social Research
• Term: since 2024
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 555455503

Established research on social networks recognizes the importance of third actors, or nodes, on social ties between individuals. For instance, Breiger's pathbreaking work on duality provides the foundations for methods considering actors as related if they are linked to the same groups or events and has been translated by subsequent work to networks with several types of persons, groups, events, or items, linked by various types of relations. Relational event models (REM, started by Butts) are statistical models for social networks given by time-ordered interaction events linking a sender to a receiver. REM assist with hypothesis testing about competing mechanisms driving the dynamics of social interaction, are increasingly applied in the empirical social sciences, and have been generalized by the applicant to relational hyperevent models (RHEM) for events simultaneously connecting more than just two actors. However, models combining the analytic capabilities and efficiency of REM with the structural generality of methods recognizing the heterogeneity of higher-order interactions are lacking. This hinders researchers in applying REM or RHEM to social network data generated by more complex interaction events without making simplifying assumptions that are often not valid in practice and that might preclude testing of relevant hypotheses. Overcoming this gap becomes pressing since, on the one hand, research in the social sciences has increasing access to empirical data with both, rich structure and fine time-granularity - on the other hand, the scarcity of adequate methods often forces researchers to assume away part of the given complexity or dynamics. This methodologically oriented project will narrow this gap between methods and the richness of theory and empirical data by developing RHEM for time-ordered events with general internal structure, represented by hyperedges containing various types of nodes, connected in a possibly non-uniform way via multiple kinds of relations. The newly developed methods will be disseminated to practitioners via the open-source software eventnet and applied in already established collaborations with social scientists from various fields. This project will make a timely contribution with widespread impact to the empirical social sciences by providing innovative and relevant social network analysis methods, applicable to a range of contemporary studies in diverse areas - narrowing the gap between methods on the one hand and the richness of theory and contemporary empirical data on the other hand - and by leveraging synergy effects with direct collaboration partners and through other social scientists adopting the novel methods in their own research.

• DFG Programme: Research Grants