Clustering Multiplex Networks in Schools Through ERGMs and Tandem Analysis

The present contribution proposes an analytical strategy designed both to identify clusters of schools characterised by distinctive local network structural configurations as well as to enhance the interpretation of the obtained results by integrating supplementary variables capturing students’ socioeconomic background and educational contexts. Using data collected from high schools in Southern Italy, the proposed approach combines results from network models and exploratory data analysis techniques. First, Exponential Random Graph Models are estimated separately for each school network, and then the resulting param-eters are used to compare relational configurations. Given the pres-ence of some outliers, in a second step, school networks are clustered by using the coordinates obtained via a robust Principal Component Analysis on the model parameters. Clusters are then interpreted in light of students’ characteristics, the scholastic context, types of rela-tionships among classmates, and educational choices after high school. The main findings reveal distinct groups, each characterised by relational patterns—from highly centralised and hierarchical to locally cohesive or expansive patterns—shaped by the interplay of peer network structures, individual characteristics, and scholastic contexts.