Spectral methods for principal components analysis of event-related brain potentials.

Principal components analysis has been a widely used method for the analysis of event-related, electrical brain potentials (ERPs). Recent emphasis has been placed on measuring the topography of ERPs, as derived from the instantaneous measurements from multiple locations, and on defining diagnostic differences in ERPs among various clinical populations. One goal of the present paper is to discuss inherent difficulties in utilizing PCA as an analytical technique in multiple location and multiple group studies. Another goal is to demonstrate the utility of spectral analysis and its equivalency to PCA when the signal imbedded in stationary noise model is used. Spectral analysis readily permits analysis of multiple lead multiple group studies.