psychometric networks
Definition
Psychometric networks refers to a modeling approach in which psychological constructs are represented as webs of pairwise associations among variables, where each variable constitutes a node and each association between variables constitutes an edge. Edges are typically estimated as partial correlations, meaning each connection reflects the unique association between two variables after controlling for all others in the system. This framework has been applied to constructs ranging from negative affect and personality to psychopathology and political attitudes. A persistent methodological challenge is that, unlike common factor models, psychometric network models do not reduce a construct to a single variable, which complicates efforts to model associations between a network-instantiated construct and external psychological variables. Research evaluating five network-derived summary metrics across four intensive longitudinal datasets found that an individual's average value of the most frequently central node was the most consistent predictor of outcomes such as life satisfaction, while network-structure variables including density, global strength, and modularity generally failed to reach significance.
Sources: Johal & Rhemtulla (2024)
Related Terms
- prediction (1 shared article)
- intensive longitudinal data (1 shared article)
Applications
Psychometric Networks and Centrality
Centrality coefficients identify which nodes are most influential within a network based on their connections to other nodes, and the most central node has been used as a summary variable for prediction. In eating disorder research, the value of the most central node has been linked to treatment response for anorexia nervosa, level of clinical impairment, and depression, while in obsessive-compulsive disorder research, the two most central symptom nodes were each associated with participants' depression and anxiety scores.
Sources: Johal & Rhemtulla (2024)
Psychometric Networks and Intensive Longitudinal Data
Intensive longitudinal data permit the estimation of individual-level network models, enabling network structure variables such as density and global strength to vary across persons rather than remaining fixed at the group level. This individual-network approach was central to evaluating which network-derived metrics carried predictive utility across four empirical datasets, where network-informed variables outperformed structural indices.
Sources: Johal & Rhemtulla (2024)
Psychometric Networks and Network Density
Network density, defined as the proportion of non-zero edges present among all possible connections in a network, has been examined as an indicator of a symptom network's capacity to spread activation. Studies comparing patient groups found higher density in individuals who persisted in a major depressive disorder diagnosis following treatment, though density did not emerge as a significant predictor of psychological outcomes in analyses using individual network models across intensive longitudinal datasets.
Sources: Johal & Rhemtulla (2024)
