network models
Definition
Network models refers to an analytical approach that represents psychological constructs by modeling direct relations between individual observed variables (such as symptoms, attitudes, or behaviors) rather than through latent variables underlying those items. The core feature of network models is that these lower-level components are conceptualized as directly influencing one another over time, forming interconnected dynamical systems that may result in stable states (e.g., a depressive episode or high neuroticism). The approach uses regularized regression techniques and structural equation modeling to estimate conditional associations or auto-regressive and cross-lagged pathways among individual items, revealing bridge nodes and specific pathways that latent variable models may overlook.
Sources: Wysocki et al. (2025)
Related Terms
Applications
Network Models and Longitudinal Panel Data
Network models have been extended to longitudinal panel designs through cross-lagged panel network (CLPN) models, which estimate item-level predictive relationships over multiple discrete measurement occasions separated by months or years. This approach combines regularized regression and structural equation modeling to reveal how individual components of psychological constructs predict each other over time, moving beyond traditional latent variable cross-lagged panel models.
Sources: Wysocki et al. (2025)
Network Models and Latent Variable Models
Network models are presented as an alternative to latent variable models for representing psychological constructs, differing fundamentally in that relations are modeled at the level of individual items and behaviors rather than at the level of latent constructs underlying those items. Network modeling approaches are most conceptually appropriate when empirical evidence or theory indicates that covariance between items is better explained by direct relations (e.g., one symptom causing another) rather than by a shared latent variable.
Sources: Wysocki et al. (2025)
