A State-Space Framework for Continuous EEG Spectral Dynamics

Google map

Electroencephalography (EEG) is commonly analyzed using predefined frequency bands (e.g., alpha: 8–12 Hz) to summarize neural activity. These band-based measures are often averaged into a single value per participant, collapsing time-varying neural activity into a static summary. While widely used, this approach assumes that frequency structure is stable across individuals and over time. This assumption may not reflect how neural activity is actually expressed, thereby obscuring meaningful within-person dynamics. In this talk, I propose a framework that reconceptualizes EEG frequency analysis as both a measurement and modeling problem. Drawing on principles of idiographic filtering, I argue that latent neural processes may be invariant across individuals, even if they appear at different frequencies. To explore this idea, I outline a two-stage analytic framework. First, person-specific factor analysis (p-technique) is used to define latent oscillatory components while allowing frequency structure to vary across individuals. Second, state-space modeling is proposed to characterize how these latent processes evolve over time. Future work will involve simulation-based validation and application of the full framework to EEG data collected during an emotion-induction task paradigm. This approach aims to move beyond static frequency band summaries toward a more precise characterization of dynamic neural processes, with potential implications for studying heterogeneous phenomena such as oscillatory dynamics in psychopathology.