Nodal heterogeneity may induce ghost triadic effects in relational event models

Abstract

Temporal network data often encode time-stamped interaction events between senders and receivers, such as co-authoring a scientific article or sending an email. A number of relational event frameworks have been proposed to address specific issues raised by modelling time-stamped data with complex temporal and spatial dependencies. These models attempt to quantify how individuals' behaviour, external factors and interaction with other individuals change the network structure over time. It is often of interest to determine whether changes in the network can be attributed to endogenous mechanisms reflecting natural relational tendencies, such as reciprocity or triadic effects, with the latter thought to represent the inherent complexity of human interaction.The propensity to form (or receive) ties can also be related to individual actors' attributes. Nodal heterogeneity in the network is often modelled by including actor-specific or dyadic covariates, such as age, gender, shared neighbourhood, etc. However, capturing personality traits such as popularity or expansiveness is difficult, if not impossible. A failure to account for unobserved heterogeneity may confound the substantive effect of key variables of interest. This research shows how node level popularity in terms of sender and receiver effects may mask ghost triadic effects. These results suggest that unobserved nodal heterogeneity plays a substantial role in REM estimation procedure and influences the conclusions drawn from real-world networks.