The GLM-spectrum: A multilevel framework for spectrum analysis with covariate and confound modelling

Area of Science:

• Neuroscience
• Signal Processing
• Statistical Modeling

Background:

• Traditional spectral estimation in electrophysiology relies on time-averaged methods established in the 1960s.
• Significant advancements in regression modeling and statistics have occurred since then, offering potential for improved analytical techniques.

Purpose of the Study:

• To introduce the General Linear Model (GLM) Spectrum, a new framework for spectral estimation in electrophysiological data.
• To demonstrate the benefits of reframing spectral estimation as a multiple regression problem.

Main Methods:

• The proposed GLM Spectrum reframes time-averaged spectral estimation as a multiple regression problem.
• This approach enables confound modeling, hierarchical modeling, and non-parametric significance testing.
• Applied to electroencephalography (EEG) data from eyes-open and eyes-closed resting states, including group-level age differences.

Main Results:

• The GLM Spectrum successfully models different conditions (eyes-open vs. eyes-closed) and quantifies their differences.
• Denoising is achieved through confound regression within a single step.
• The method is scalable from single-channel to whole-head recordings and group-level analyses.
• Model-projected spectra offer intuitive visualization for within- and between-subject contrasts and interactions.

Conclusions:

• The GLM Spectrum provides a flexible and robust framework for multilevel analysis of power spectra.
• It allows for adaptive modeling of covariates and confounds, enhancing the rigor of electrophysiological data analysis.
• This method facilitates comprehensive analysis of spectral dynamics, including complex contrasts and interactions.