Regression-based ERP/ERF analysis

rERP/rERF analysis is a technique for using regression to estimate ERP/ERF waveforms from brainwave data recorded using EEG or MEG.

Traditional ERP/ERF analysis is designed for comparing categorical conditions (expected/unexpected, attended/unattended, etc.), and runs into problems if you have an experiment with continuous covariates, correlations between your covariates (partial confounding), or present stimuli in quick enough succession to create overlap between their brain responses. rERP/rERF analysis includes ERP/ERF analysis as a special case, but it extends it to handle experiments containing categorical covariates, continuous covariates, or both; partial confounding; non-linear effects; overlapping responses to distinct events; and any combination of the above.

rERP publications (see also references therein):
  • Smith, N. J., & Kutas, M. (2015). Regression-based estimation of ERP waveforms: I. The rERP framework. Psychophysiology, 52(2), 157-168. (PDF)
  • Smith, N. J., & Kutas, M. (2015). Regression-based estimation of ERP waveforms: II. Non-linear effects, overlap correction, and practical considerations. Psychophysiology, 52(2), 169-189. (PDF (including supplementary information))
  • Burns, M. D., Bigdely-Shamlo, N., Smith, N. J., Kreutz-Delgado, K., & Makeig, S. (2013). Comparison of Averaging and Regression Techniques for Estimating Event Related Potentials. In IEEE Engineering in Biology and Medicine Conference, Osaka, Japan. (PDF)
Relevant software:
  • rERPy: Python, can calculate arbitrary rERP/rERF designs with or without overlap correction, powerful tools for describing and selecting events and creating design matrices. Not yet documented. Used for all examples in the Smith & Kutas papers above.
  • LIMO: EEGLAB plugin, may not follow all recommendations in the above manuscripts by default but with some work I believe should be able to perform rERP/rERF regression analyses using arbitrary design matrices. No support for overlap correction. Also has powerful tools for between-subjects statistical analysis that I don't understand in detail so can't comment on.
  • rERP toolbox: EEGLAB plugin, limited tools for defining design matrices, can compute both rERP/rERFs and rERSPs with overlap correction. Used for the computations in Burns et al (2013).
Feedback/comments/questions:
To Nathaniel J. Smith