[Eeglablist] ERSP and ERP analyse using MPT clustering method

Tarik S Bel-Bahar tarikbelbahar at gmail.com
Mon May 11 16:10:57 PDT 2015


Hello Mohammed, hoping the brief thoughts below help a little. ~Best wishes.

1. There should be a relationship between ERP and ERSP, in the sense that
chances in spectral dynamics "create" ERPS, thus it's probable that using
both features should give more valid clusters.
However if it's suggested to run MPT on one feature at a time, please do so.
2. I would say try both features together for clustering, and also try them
separately for two different clustering runs. You may empirically show the
difference, and that would be something useful for the field.
If there's no difference that also useful to note [let the list know if
possible]. If you haven't tried things with the traditional k-means
clustering in matlab, it would be interesting to know if you see any
difference with that method compered to MPT clustering.
3. It matters to some degree whether the condition effects you are
interested in show up more (or less) in the ICAs using different filtering
approaches, or in the clustering with one or two features.
4. I don't think you'll have a problem with reporting on comparisons of the
locations of spatial clusters when the data going into the ICAs has been
filtered differently,
as long as you describe appropriate caveats. If you do clustering  in
several ways then you could  report whether or not there is a difference in
localization across the techniques you choose.





On Mon, May 11, 2015 at 5:46 AM, Mohammed Jarjees <
m.jarjees.1 at research.gla.ac.uk> wrote:

> Dears,
>
>
>
> I want to cluster my 61 electrodes EEG data for three groups of people
> each group has 10 subjects and each subjects has three conditions.
>
> I would like to analyse ERSP and ERP and to use MPT clustering method. MPT
> suggest to use one feature at the time, so I will create separate studies
> for ERSP and ERP. I would like to be able to compare clustering of ERSP and
> of ERP as these two variables might not necessarily have the same clusters.
>
>
>
>
>
> My questions are:
>
> 1. Should I select the same ICAs for ERP and ERSP?  Though this might
> sound a common sense, by visual inspection of ICAs I've noticed that some
> ICs have very strong ERP while not much is going on in ERSP and vice-versa,
> some have very little ERP with strong ERSP. Therefore taking common
> components for both might blur  results for both ERP and ERSP.
>
> 2. In case I should not use the same components, may I produce two ICs
> sets, one filtered at 0.1Hz for ERP and the other filtered with 1Hz for
> ERSP. That is because filtering with 1Hz for ERSP really helps removing a
> baseline drift and I get clearer ICs then when filtering with 0.1 Hz (which
> is necessary should I analyse ERP)?
>
> Would I still be able to compare spatial locations of ERP and ERSP
> clusters?
>
>
>
> Thank you very much for your help.
>
> Best regards,
>
> Mohammed
>
>
>
> _______________________________________________
> 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
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20150511/cac461df/attachment-0001.html>


More information about the eeglablist mailing list