Technical Program

Paper Detail

Paper: PS-2A.10
Session: Poster Session 2A
Location: H Lichthof
Session Time: Sunday, September 15, 17:15 - 20:15
Presentation Time:Sunday, September 15, 17:15 - 20:15
Presentation: Poster
Publication: 2019 Conference on Cognitive Computational Neuroscience, 13-16 September 2019, Berlin, Germany
Paper Title: Unmixed: Linear Mixed Models combined with Overlap Correction for M/EEG analyses. An Extension to the unfold Toolbox
Manuscript:  Click here to view manuscript
License: Creative Commons License
This work is licensed under a Creative Commons Attribution 3.0 Unported License.
Authors: Benedikt V. Ehinger, Radboud University, Netherlands
Abstract: Linear mixed models (LMMs) offer several benefits over traditional two-stage analysis methods common in EEG analysis: Higher power to detect effects, partial pooling with noisy data and the possibility to account for both subject and item effects. LMMs come at the price of increased computational cost, up to now making them incompatible to use in natural experiments that require time-resolved deconvolution methods of continuous EEG data. Here, I present unmixed an extension to the open source unfold-toolbox, allowing to fit LMMs and GAMMs to rERP (regression ERPs) using extended Wilkinson formulas. Unmixed supports mixed modelling of overlapping events and non-linear effects. It offers several different optimizers, Walds t-tests and likelihood ratio model comparison tests for statistical analysis, and Benjamini-Hochberg FDR for multiple comparison correction. This technique is promising for population where extensive data collection is not possible, e.g. infants or clinical populations.