Simulating Relational Event Histories: Why and How

TL;DR

This work addresses the need to study social interaction dynamics through time-stamped relational events by introducing two simulation frameworks, REM and DyNAM, and the remulate R package for flexible, memory-aware simulation of relational histories. It demonstrates how simulations can be used for absolute model assessment, theory development (via Optimal Distinctiveness Theory), and planning network interventions, illustrated through gang-violence fit checks, theory-driven group formation, and intervention persistence and targeting experiments. The proposed approach enables researchers to inspect how well models reproduce temporal and structural network characteristics, test boundary conditions of theories, and compare intervention scenarios in a risk-free in silico environment, thereby guiding empirical study design and policy planning. Overall, the paper highlights the practical impact of simulation-based temporal network analysis for understanding, predicting, and influencing social interaction dynamics in real-world networks.

Abstract

Many important social phenomena are characterized by repeated interactions among individuals over time such as email exchanges in an organization or face-to-face interactions in a classroom. To understand the underlying mechanisms of social interaction dynamics, statistical simulation techniques of longitudinal network data on a fine temporal granularity are crucially important. This paper makes two contributions to the field. First, we present statistical frameworks to simulate relational event networks under dyadic and actor-oriented relational event models which are implemented in a new R package 'remulate'. Second, we explain how the simulation framework can be used to address challenging problems in temporal social network analysis, such as model fit assessment, theory building, network intervention planning, making predictions, understanding the impact of network structures, to name a few. This is shown in three extensive case studies. In the first study, it is elaborated why simulation-based techniques are crucial for relational event model assessment which is illustrated for a network of criminal gangs. In the second study, it is shown how simulation techniques are important when building and extending theories about social phenomena which is illustrated via optimal distinctiveness theory. In the third study, we demonstrate how simulation techniques contribute to a better understanding of the longevity and the potential effect sizes of network interventions. Through these case studies and software, researchers will be able to better understand social interaction dynamics using relational event data from real-life networks.

Figures (6)

• Figure 1: Goodness-of-fit comparison. Each figure plots the values of the fit indices over 100 simulated sequences and the observed data. In figures (a-f) the red crosses represent the observed data and the boxplots depict the distribution of the fit indices across simulated networks. In figures (g-h) the dotted red line represents the observed data.
• Figure 2: Simulated Network plots for different optimal distinctiveness $d^*$ values. Square nodes represent nodes with $z=1$ and circular nodes represent nodes with attribute value $z=2$. The colours indicate the emergent groups to which the nodes were assigned based on the community detection algorithm. The size of nodes corresponds to the degree of the actors, and the width of the edges corresponds to the total volume of events exchanged in either direction between the two nodes.
• Figure 3: (a) Average distinctiveness of actors, (b) Average modularity computed based on the partitioning of nodes by actor attribute and, (c) Average modularity computed based on the partitioning of nodes by the detected groups at the end of 50 simulation runs for each value of optimal distinctiveness $d^* \in [0,1]$ and for each proportion of minority attribute.
• Figure 4: Diagrammatic representation of two interventions
• Figure 5: Median proportion of inter-departmental events simulated after an artificial intervention. The vertical dotted lines indicate the the intervention period. Top panel shows the simulations for an intervention lasting 4 weeks i.e 28 days in duration. Middle panel with duration 6 weeks i.e 42 days and Bottom panel with 8 weeks i.e 56 days.