[Eeglablist] Concatenate data across visits prior to ICA?

[Scott Burwell]

Others may want to chime in, but my two cents follow:
1. I think that concatenating runs (e.g., tasks) within a given recording
session for the purposes of ICA is plenty defensible, especially if you
want to compare source activities across resting-state, task, etc. This of
course assumes that the EEG cap placement has not changed within that
session. One thing you might want to consider, if using AMICA, if you want
to allow for multiple models? Presumably, the impact of certain groups of
effective EEG source generators (artifacts and "true" brain sources) change
across different tasks / states, which could be reflected in different
AMICA models (e.g., Hsu, 2018; NeuroImage).
2. I do NOT think that concatenating data across different sessions
would be defensible, unless somehow modeling for differences in cap
placement. One of the main assumptions of an ICA model is that the
underlying source distributions do not change. If your cap placement
changes (e.g., by a few mm in AP and a few mm RL, or more), the
channel-to-source distributions change, causing problems with the model fit
for either data session.

Best,
Scott

On Mon, Mar 23, 2020 at 9:56 AM Coombes, Stephen A <scoombes at ufl.edu> wrote:

> Dear EEG list members,
>
> We have EEG data from a cohort of subjects who come into the lab for 3
> visits spaced over a 6 month period. It is a repeated measures design with
> visit being the only variable. We have resting data and data during a task.
> For each data set (i.e., resting & task based), should we concatenate the
> data for each subject across visit before running ICA, and if so why?
>
> For the task based analysis we will run ICA then DIPFIT and then MPA. We
> are still exploring different options for the resting state pipeline once
> ICA and artifact removal is complete.
>
> Any advice greatly appreciated. Hope everyone is safe and well out there.
>
> Many thanks,
> Steve
>
> __________
> Stephen A. Coombes, Ph.D.
> Laboratory for Rehabilitation Neuroscience
> University of Florida
> Phone: 352.294.1768
> Website: lrnlab.org<http://www.lrnlab.org/>
>
> _______________________________________________
> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> To unsubscribe, send an empty email to
> eeglablist-unsubscribe at sccn.ucsd.edu
> For digest mode, send an email with the subject "set digest mime" to
> eeglablist-request at sccn.ucsd.edu
>


-- 
Scott J. Burwell, PhD
NIDA T32 Postdoctoral Research Fellow
Department of Psychiatry & Behavioral Sciences
University of Minnesota, Minneapolis, MN
burwell at umn.edu
github.com/sjburwell