Group-Level EEG-Processing Pipeline for Flexible Single Trial-Based Analyses Including Linear Mixed Models

• Published in: Frontiers in neuroscience Vol. 12; p. 48
• Main Authors: Frömer, Romy, Maier, Martin, Abdel Rahman, Rasha
• Format: Journal Article
• Language: English
• Published: Switzerland Frontiers Research Foundation 06.02.2018; Frontiers Media S.A
• Subjects: cluster-based permutation tests; Cognitive ability; Economic models; EEG; EEGLab; Estimates; Event-related potentials; Familiarity; Linear mixed models; Neuroscience; processing pipeline; Variables; Visual perception; Germany; Linear mixed models; processing pipeline; EEGLab; cluster-based permutation tests; EEG
• ISSN: 1662-453X; 1662-4548; 1662-453X
• DOI: 10.3389/fnins.2018.00048

Here we present an application of an EEG processing pipeline customizing EEGLAB and FieldTrip functions, specifically optimized to flexibly analyze EEG data based on single trial information. The key component of our approach is to create a comprehensive 3-D EEG data structure including all trials and all participants maintaining the original order of recording. This allows straightforward access to subsets of the data based on any information available in a behavioral data structure matched with the EEG data (experimental conditions, but also performance indicators, such accuracy or RTs of single trials). In the present study we exploit this structure to compute linear mixed models (LMMs, using lmer in R) including random intercepts and slopes for items. This information can easily be read out from the matched behavioral data, whereas it might not be accessible in traditional ERP approaches without substantial effort. We further provide easily adaptable scripts for performing cluster-based permutation tests (as implemented in FieldTrip), as a more robust alternative to traditional omnibus ANOVAs. Our approach is particularly advantageous for data with parametric within-subject covariates (e.g., performance) and/or multiple complex stimuli (such as words, faces or objects) that vary in features affecting cognitive processes and ERPs (such as word frequency, salience or familiarity), which are sometimes hard to control experimentally or might themselves constitute variables of interest. The present dataset was recorded from 40 participants who performed a visual search task on previously unfamiliar objects, presented either visually intact or blurred. MATLAB as well as R scripts are provided that can be adapted to different datasets.