Relational Hypervents Models for Co-authorship Networks: The Case of the Italian Academic Statisticians

In the last year, there has been a growing emphasis on the significance of scientific collaboration. Several studies have analyzed co-authorship networks as a convenient tool to investigate collaboration among scientists across different disciplines and over time. Co-authorship networks among scientists can be represented by hypergraphs, where each scientific paper is the corresponding hyperevent. To account for this scenario and evaluate the evolution of the co-authorship network over time, Relational Hypervent Models (RHEM) have been recently proposed. Such models can handle time-varying data and events that occur in (weighted) hyperedges. They represent a suitable class of models for co-authorship data as they can model events composed of any authors, analyze polyadic settings, and consider sub-group persistence over time. This paper aims to apply such models to analyse the evolution of collaboration among the Italian academic statisticians observed over 10 years, from 2012 to 2022, including their external co-authors. Bibliographic information has been retrieved from the Scopus database. Different model specifications are proposed to evaluate the tendency to maintain collaboration among authors over time (familiarity effects) and/or the tendency to create new collaborations with authors who co-authored with common groups of authors (closure effects).