personalized models; moderation analysis
Definition
Personalized models, moderation analysis refers to the integration of idiographic, person-specific network modeling with formal moderation techniques to identify which contextual factors drive changes in an individual's psychological system over time. Standard idiographic network approaches, built on vector autoregressive models, assume that connections among variables and their mean levels remain stationary, leaving the causes of network change unaccounted for. The fixed moderated time series model addresses this limitation by allowing all parameters of the VAR model, including the innovation structure and mean levels, to be moderated by external variables such as social interaction or sleep quality. Applied to patients receiving treatment for depression, this approach can reveal whether a binary or continuous moderator shifts the temporal dynamics between symptoms, the variance attributable to unobserved factors, or both. The accompanying R function mod_mlvar implements estimation via a two-step Bayesian procedure, making the method accessible for clinical research seeking to explain, rather than merely detect, change in personalized psychological networks.
Sources: Bringmann et al. (2024)
Related Terms
- networks (1 shared article)
- time series (1 shared article)
- vector autoregressive model (1 shared article)
Applications
Personalized Models; Moderation Analysis and Idiographic Psychological Networks
Idiographic psychological networks represent associations among an individual's symptoms, emotions, and behaviors across time using intensive longitudinal data, but in their standard form they do not model why those associations change. The fixed moderated time series model extends idiographic networks by treating network parameters as functions of moderating variables, enabling clinicians and researchers to identify which contextual factors, such as being alone versus with others, alter the structure of a patient's network.
Sources: Bringmann et al. (2024)
Personalized Models; Moderation Analysis and Vector Autoregressive Models
Vector autoregressive models form the statistical foundation of personalized psychological networks, capturing lagged relationships among variables at successive time points. Existing VAR-based time-varying approaches identify when network connections change but cannot attribute those changes to specific moderators, and they do not moderate the innovation structure. The fixed moderated time series model resolves this by allowing every component of the VAR, including the error variance and mean levels, to be moderated, estimated within a state space framework.
Sources: Bringmann et al. (2024)
Personalized Models; Moderation Analysis and Clinical Intervention
Identifying moderators of network change carries direct implications for treatment, because it specifies which factors a clinician might target to shift a patient's psychological system toward a healthier state. In the empirical examples presented, testing whether sleep quality moderates connections in a depression patient's emotion network provides guidance on whether improving sleep could alter the dynamics between symptoms such as feeling down and cheerfulness.
Sources: Bringmann et al. (2024)
