[Eeglablist] Individual participant contributions to source-space clusters

[Makoto Miyakoshi]

Dear Matt,

*> Questions*

Is ICA appropriate for what I want to achieve, or do you suggest other
methods of source localisation?


Technically speaking, ICA is not itself a 3-D source localizer.

You can use LORETA on ICA-derived scalp topography, for example.


For your purpose, ICA can be used. I mean, I can't think of any case where
ICA cannot be used... could it be something like, for example, data in
which all channels showed complete Gaussian distributions?


> Do you have a strategy on how to approach the EEGLAB data-structures on a
participant level to access the information I require?


Unfortunately, STUDY structure is complicated. When I came to SCCN I could
barely retrieve individual subject data from STUDY strucuture. I could
still figure it out, without knowing this page, and it took me a few days.
At the bottom of this wiki page, it says 'This apparently complex scheme...'


https://sccn.ucsd.edu/wiki/Chapter_07:_EEGLAB_Study_Data_Structures#Understanding_the_.sets.2C_.comps.2C_.setinds.2C_.allinds_substructures_for_STUDY_clusters


So your good strategy would be to read the wiki page and efficiently figure
out how to retrieve individual data.


Makoto

On Mon, Sep 19, 2016 at 10:47 PM, Matt Gerhold <matt.gerhold at gmail.com>
wrote:

> ICA gurus:
>
> I would be very grateful if you can assist me in understanding some of the
> methods and data structures from your toolbox. Specifically, I would like
> to gain some clarity regarding working with clusters of independent
> components in the source-space, wherein individual participants contribute
> source-localised components to a group level cluster.
>
>
>
> *What I have done and what I have observed*
>
> I have pre-processed the data (including applying ICA), performed dipole
> source localisation, and clustered the components. The resultant group
> level clusters make sense: I am using a task that requires that
> participants respond with a button press, or alternatively withhold a
> button press when cued to do so. I have observed upper alpha (*rolandic
> mu*) desynchronization in the component cluster localised to the motor
> region contralateral to the button-pressing finger. In addition, other such
> event-related spectral features suggest that the clustering and dipole
> localisation have been successful.
>
> When I inspect the clustered components, I have observed that some
> clusters may contain two components from the same participant—the two
> components have a similar scalp distribution and spectra leading to similar
> clustering. In certain instances, some group level clusters may not have
> contributions from one or two participants within the group. Therefore, on
> the group level the time-frequency data looks good; however, the observed
> group level dynamics do not easily translate into individual participant
> measures that can be tabulated for a more rigorous statistical analysis.
>
>
>
> *What I am trying to achieve*
>
> I need to check for confounding (a covariate analysis) and possible
> mediating effects using a general linear model framework. For this, I have
> a collection of measurements for each participant and require a spectral
> feature representative of the group level cluster for each participant in
> order to perform my analysis. For example, alpha power over the
> contralateral motor cortex in a post-stimulus onset window would be a
> spectral feature; I would potentially use age-at-recording as a covariate
> in my analysis to see if maturational effects explain variation in spectral
> power over-and-above the experimental manipulation(s).
>
>
>
> *Questions*
>
> Is ICA appropriate for what I want to achieve, or do you suggest other
> methods of source localisation?
>
> Do you have a strategy on how to approach the EEGLAB data-structures on a
> participant level to access the information I require?
>
>
>
> Any advice related to the above questions would be greatly appreciated.
> Many thanks in advance.
>
> Kind Regards,
>
> Matthew
>
> _______________________________________________
> 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
>



-- 
Makoto Miyakoshi
Swartz Center for Computational Neuroscience
Institute for Neural Computation, University of California San Diego
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20161017/b962fbab/attachment.html>