Familiarity and Liking in Classroom Networks: A Multilayer Analysis of Rated Peer Relations

Many social network datasets consist of ordinal ratings exchanged among the same set of individuals on multiple relational dimensions. Such data are common in educational and behavioral research but pose challenges due to ordinal measurement, asymmetric sender and receiver roles, and dependence within and across relations. We analyze classroom network data from Taiwanese senior high schools, in which students provided ratings of familiarity and liking, to examine relational dimensionality, reciprocity, cross-relational overlap, and individual characteristics. To this end, we develop and apply a latent-space rated-network model that accounts for ordinal rating scales, sender and receiver heterogeneity, and higher-order interaction structure. The model allows familiarity and liking to be treated either as a single relational dimension or as distinct but interdependent layers and decomposes within- and cross-layer dependence into interpretable, role-specific components linked to classical network summaries, such as reciprocity and multiplexity. In application, the model shows that high descriptive overlap between familiarity and liking does not imply a common generative process. Latent interaction patterns are shared across relations, but differences emerge in dispersion of latent positions, particularly by gender. Moreover, familiarity ratings are associated with both sender and receiver personality traits, whereas liking ratings are primarily associated with receiver traits.