@misc{BridwellCavanaghCollinsetal.2018,
  author    = {Bridwell, David A. and Cavanagh, James F. and Collins, Anne G. E. and Nunez, Michael D. and Srinivasan, Ramesh and Stober, Sebastian and Calhoun, Vince D.},
  title     = {Moving beyond ERP components},
  series = {Postprints der Universit{\"a}t Potsdam : Humanwissenschaftliche Reihe},
  journal   = {Postprints der Universit{\"a}t Potsdam : Humanwissenschaftliche Reihe},
  number    = {656},
  issn      = {1866-8364},
  doi       = {10.25932/publishup-45966},
  url       = {http://nbn-resolving.de/urn:nbn:de:kobv:517-opus4-459667},
  pages     = {19},
  year      = {2018},
  abstract  = {Relationships between neuroimaging measures and behavior provide important clues about brain function and cognition in healthy and clinical populations. While electroencephalography (EEG) provides a portable, low cost measure of brain dynamics, it has been somewhat underrepresented in the emerging field of model-based inference. We seek to address this gap in this article by highlighting the utility of linking EEG and behavior, with an emphasis on approaches for EEG analysis that move beyond focusing on peaks or "components" derived from averaging EEG responses across trials and subjects (generating the event-related potential, ERP). First, we review methods for deriving features from EEG in order to enhance the signal within single-trials. These methods include filtering based on user-defined features (i.e., frequency decomposition, time-frequency decomposition), filtering based on data-driven properties (i.e., blind source separation, BSS), and generating more abstract representations of data (e.g., using deep learning). We then review cognitive models which extract latent variables from experimental tasks, including the drift diffusion model (DDM) and reinforcement learning (RL) approaches. Next, we discuss ways to access associations among these measures, including statistical models, data-driven joint models and cognitive joint modeling using hierarchical Bayesian models (HBMs). We think that these methodological tools are likely to contribute to theoretical advancements, and will help inform our understandings of brain dynamics that contribute to moment-to-moment cognitive function.},
  language  = {en}
}
@article{BuerkiFoschiniFrossardRenaud2018,
  author    = {B{\"u}rki-Foschini, Audrey Damaris and Frossard, Jaromil and Renaud, Olivier},
  title     = {Accounting for stimulus and participant effects in event-related potential analyses to increase the replicability of studies},
  series = {Journal of neuroscience methods},
  volume    = {309},
  journal   = {Journal of neuroscience methods},
  publisher = {Elsevier},
  address   = {Amsterdam},
  issn      = {0165-0270},
  doi       = {10.1016/j.jneumeth.2018.09.016},
  pages     = {218 -- 227},
  year      = {2018},
  abstract  = {Background: Event-related potentials (ERPs) are increasingly used in cognitive science. With their high temporal resolution, they offer a unique window into cognitive processes and their time course. In this paper, we focus on ERP experiments whose designs involve selecting participants and stimuli amongst many. Recently, Westfall et al. (2017) highlighted the drastic consequences of not considering stimuli as a random variable in fMRI studies with such designs. Most ERP studies in cognitive psychology suffer from the same drawback. New method: We advocate the use of the Quasi-F or Mixed-effects models instead of the classical ANOVA/by-participant F1 statistic to analyze ERP datasets in which the dependent variable is reduced to one measure per trial (e.g., mean amplitude). We combine Quasi-F statistic and cluster mass tests to analyze datasets with multiple measures per trial. Doing so allows us to treat stimulus as a random variable while correcting for multiple comparisons. Results: Simulations show that the use of Quasi-F statistics with cluster mass tests allows maintaining the family wise error rates close to the nominal alpha level of 0.05. Comparison with existing methods: Simulations reveal that the classical ANOVA/F1 approach has an alarming FWER, demonstrating the superiority of models that treat both participant and stimulus as random variables, like the Quasi-F approach. Conclusions: Our simulations question the validity of studies in which stimulus is not treated as a random variable. Failure to change the current standards feeds the replicability crisis.},
  language  = {en}
}