[Eeglablist] How much data is necessary for ICA?

[Scott Makeig]

Dana -

Note that performing PCA BEFORE ICA is *not* recommended (see paper linked
below) unless the data are rank-deficient, so ICA cannot decompose them --
in which case PCA is a sound way to proceed).

Artoni F, Delorme A, S Makeig. Applying dimension reduction to EEG data by
principal component analysis reduces the quality of its subsequent
independent component decomposition.
<https://sccn.ucsd.edu/~scott/pdf/Artoni_PCAICA18.pdf> *NeuroImage*
175:176–187,
2018. https://urldefense.com/v3/__https://doi.org/10.1016/j.neuroimage.2018.03.016__;!!Mih3wA!VBHJh6UIj2hBOWFJk_wxHBS2kyWcnlAXe6ZQjtCtZdC1KO9Y2_zPBgR44Szkz7bGPrQo8A$ 
<https://urldefense.com/v3/__https://www.sciencedirect.com/science/article/pii/S1053811918302143__;!!Mih3wA!VBHJh6UIj2hBOWFJk_wxHBS2kyWcnlAXe6ZQjtCtZdC1KO9Y2_zPBgR44Szkz7aoZUOazg$ >

Scott Makeig

On Mon, Apr 27, 2020 at 5:09 AM Son, D.M.E. van <
d.m.e.van.son at fsw.leidenuniv.nl> wrote:

> Dear EEGLab list members,
>
> I have a question related to ocular and artifact rejection with ICA to
> obtain ERPs. I read ICA might not be ideal if there are not enough data
> points (e.g. if you only have a short EEG recording per subject). There
> seem to be ways to cope with this, such as reducing the number of channels,
> having a higher sample rate or using PCA.
>
> We will have 20 minutes of data per subject, 128 channels and recorded the
> data with 1024 p/s. Does anyone know whether this will be enough to
> confidently apply the ICA? Is reducing the number of channels or an extra
> PCA before ICA recommended?
>
> Many thanks in advance
>
> Dana
> _______________________________________________
> 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 Makeig, Research Scientist and Director, Swartz Center for
Computational Neuroscience, Institute for Neural Computation, University of
California San Diego, La Jolla CA 92093-0559, http://sccn.ucsd.edu/~scott