network clustering
Definition
Network clustering refers to the identification of groups of nodes that are more densely interconnected with one another than with nodes outside the group, a procedure applied in both psychometric and cognitive network frameworks to reveal latent structure in complex data. In the context of Exploratory Graph Analysis, items from a psychometric scale such as the Depression Anxiety and Stress Scale are clustered according to patterns in participant ratings, yielding factors corresponding to distinct psychological dimensions including anxiety, stress, and depression. A parallel clustering procedure is applied to Textual Forma Mentis Networks, where semantic communities are extracted as groups of words sharing tighter syntactic and semantic connections within the text of questionnaire items than with words outside the community. The degree to which these two types of clusters, one derived from numerical ratings and one from the semantic structure of item wording, overlap is quantified through semantic loadings, which map how topic-based communities distribute across psychometric factors. When applied to 39,775 responses to the 42-item DASS, this approach revealed non-random correspondences between semantic communities and psychometric factors at p less than 0.001.
Sources: Stanghellini et al. (2024)
Related Terms
- network psychometrics (1 shared article)
- cognitive network science (1 shared article)
- text analysis (1 shared article)
- psychometric measurements (1 shared article)
- semantic framing (1 shared article)
Applications
Network Clustering and Exploratory Graph Analysis
Exploratory Graph Analysis applies network clustering to psychometric response data in order to determine the number and composition of psychological factors, functioning as an alternative to eigenvector-based factor analysis. In the DASS study, EGA recovered the three-dimensional structure of depression, anxiety, and stress as distinct psychometric factors.
Sources: Stanghellini et al. (2024)
Network Clustering and Semantic Communities
Semantic communities are clusters extracted from Textual Forma Mentis Networks, where words are grouped by the density of their syntactic and semantic links across the texts of psychometric items. These communities differ from psychometric factor clusters in that they arise from the meaning structure encoded by scale designers rather than from participant ratings, and semantic loadings are defined as the overlap between the two types of clusters.
Sources: Stanghellini et al. (2024)
Network Clustering and Depression, Anxiety and Stress
Network clustering of DASS item responses via EGA produced factors corresponding to emotional dysregulation, emotional exhaustion, physical distress, and tension states, mapping onto the constructs of depression, anxiety, and stress. Semantic communities identified through clustering of the DASS Textual Forma Mentis Network aligned with these same psychological dimensions in non-random ways, providing a converging characterisation of the constructs from two independent clustering procedures.
Sources: Stanghellini et al. (2024)
