EUSN 2021 Book of Abstracts. 5th European Conference on Social Networks

In recent years, the analysis of ego networks has attracted a great attention and found application in many areas of the social sciences. Several studies have pointed out the crucial role played by network characteristics (such as size and composition) in the study of social relationships and their impact on many aspects of everyday life (e.g., social support, well-being, health, and mobility). In this context, the identification of network typologies has become a crucial task and a powerful tool to capture aspects of the social space or personal community in which people are embedded. Along this direction, clustering methods have been exploited to identify and characterize existent types of ego networks. In this work, we propose a distance-based clustering procedure to identify groups of similar ego networks, which are described by a small number of compositional variables. The proposed approach is motivated by the empirical study of ego networks of contacts extracted from the latest edition of “Family and Social Subjects" (FSS) Survey conducted by the Italian National Statistical Institute in 2016. In particular, we focus on elderly respondents living alone, which can be regarded as a vulnerable category, with the aim to describe their network of contacts. As the FSS Survey is not specifically oriented to network analysis, its major limitation consists in the lack of information on alter-alter ties. Coping with these limitations, we first mine relational information in FSS data in order to derive the ego networks of respondents. Then, we develop a clustering procedure in the hierarchical framework to identify a partition of ego networks according to their composition. The proposed approach has the main advantage to be particularly suitable when the involved variables are heterogeneous, in both range and type, which can easily happen if ego networks are derived from secondary data, rather than using ad-hoc designs. We discuss the choice of a suitable dissimilarity metric and the issue of the selection of the number of clusters. The prototypical units---one for each cluster---resulting from the proposed method, enhance the cluster interpretation. Results on the analysis of ego networks for the elderly people show the suitability of the proposed procedure to investigate the existing patterns in egocentric data, especially when the data are defined over heterogeneous attributes concerning the network composition, making our approach applicable to various surveys.