Nodal Heterogeneity can Induce Ghost Triadic Effects in Relational Event Models

Area of Science:

• Network Science
• Computational Social Science
• Statistical Modeling

Background:

• Temporal network data captures interactions over time, like co-authorship or emails.
• Relational event frameworks model temporal dependencies and network dynamics.
• Understanding endogenous mechanisms (reciprocity, triadic effects) and actor attributes is crucial.

Purpose of the Study:

• To address issues of nodal heterogeneity in temporal network analysis.
• To demonstrate how unobserved sender and receiver effects can create spurious triadic effects.
• To propose a novel random-effect extension for relational event models.

Main Methods:

• Developed a random-effect extension of the relational event model.
• Compared the proposed model against traditional approaches like in-degree and out-degree statistics.
• Investigated the impact of unobserved nodal heterogeneity on network dynamics.

Main Results:

• Failing to account for sender and receiver effects can induce 'ghost' triadic effects.
• The proposed random-effect model effectively resolves issues caused by nodal heterogeneity.
• The random-effect extension outperforms traditional methods in capturing network dynamics.

Conclusions:

• Including random effects in relational event models is essential for accurately capturing network dynamics.
• This approach resolves the violation of the hierarchy principle caused by insufficient information on nodal heterogeneity.
• The proposed method offers a robust solution for analyzing complex temporal network data.