[Eeglablist] selectICsByCluster -- two within-subject conditions

Tarik S Bel-Bahar tarikbelbahar at gmail.com
Sun Oct 8 16:19:31 PDT 2017


Hello Haiyang,

The recommendations from Makoto and his Std_selectICsByCluster
<https://sccn.ucsd.edu/wiki/Std_selectICsByCluster> tool are both pretty
cool and useful, and kudos to you for giving them all a try.
here's a few last points below, I hope they are helpful to you a little,
they are only simple suggestions or reminders. You already seem like a
strong eeglab user, so keep trying, you should find success slowly but
surely. Hopefully we'll hear back from Mikayoshi-san when he has time, and
he may provide useful clarifications or fixes.

1. It's possible that the tool is not fully catching all problems, and you
will probably have to wait until he has time to debug/fix it. That being
said, it might be something you are doing wrong, but we can't know unless
we had a lot of documentation on the tool.

2. Good work on doing all those steps with script. I would ask if you ever
did it with the GUI, and what did you find when you did things that way.
It's usually helpful to try things first from the gui, and use eegh to
understand the steps that are being done. It's very easy to miss something,
or have an expectation that script-only will work out of the box. Generally
speaking, script-only solutions will tend to have more bugs, and especially
when using "extra tools".

3. I would recommend trying the normal study gubased clustering and pruning
process, if you have the time and inclination.

4. I would also recommend simply getting the IC's to be dropped per
single-subject, and then looping through the single-subject files (not in
study) and removing IC's that you don't want to keep from each
single-subject file. Then load up these pruned single-subject files into a
new study and see how far you can get.

5. Last, you must take a closer look at the STUDY documentation and
understand whether or not you are making a problem by asking it to do a
design it cannot do. In essence, try a simpler design and/or be prepared to
build your own study design by using single-subject data and using your own
personally developed code (e.g., create matlab hypercube of all metrics you
want, for each subject/condition/group, and apply stats functions to parts
of the hypercube).


​Hi

Thanks.


I'm following makoto's preprocessing pipeline (data are not epoched before
AMICA).


For this part, I'm trying to reject non-EEG independent components via
study design by following makoto (https://sccn.ucsd.edu/wiki/
Std_selectICsByCluster).


I use scripts to create and load the study, precompute specturm, cluster
all  ICs jsut by using spectra (then run 'eeglab redraw'). After that, I
use the GUI to create the same number of clusters, plot spectra to write
down the good ICs, and then launch the std_selectICsByCluster() plugin and
enter the good cluster indices to save the .set files with only good ICs.
During these, no error comes out.


However, (1) I found the produced .set files with only good ICs have the
same sizes as the old (original) ones. Should the sizes of  .set files with
only good ICs are smaller than the original ones since some (bad) ICs have
beed deleted?

(2), After the processing mentioned above, I tried to run
std_plot_selectICsByCluster via GUI for the .set files with only good ICs,
an error says 'Reference to non-existent field 'selectICsByCluster'.'


In total, I'm not sure if I reject the non-EEG ICs as a group level filter
correctly.​ BTW, my experiment is a 2*4*2 winthin-subject design.


Thanks again.


Haiyang



------------------------------
*From:* Tarik S Bel-Bahar <tarikbelbahar at gmail.com>
*Sent:* Monday, 9 October 2017 7:46
*To:* Haiyang Jin
*Cc:* eeglablist at sccn.ucsd.edu
*Subject:* Re: [Eeglablist] selectICsByCluster -- two within-subject
conditions

Hello Haiyang,

Overall i believe you are probably correct, but Makoto can get back to you
later.
Can you provide more detailed information here?
Which function are you referring to, and are you accessing it via the STUDY
GUI or via a script only ?
What have you tried that does work ?
You can also review STUDY documentation online to determine if STUDY is
limited in this way.



On Thu, Oct 5, 2017 at 2:55 AM, Haiyang Jin <haiyang.jin at auckland.ac.nz>
 wrote:

> Hi Makoto
>
>
> Just to confirm that two within-subject conditions are not supported by
> selectICsByCluster.
>
>
> For example, a experiment design with stimulus (face vs house) and
> orientation (upright and inverted) is not support by the
> selectICsByCluster.
>
>
> Kind regards
>
>
> Haiyang
>
>
> Haiyang Jin
> Ph.D Candidate
> Visual Cognition Lab
> Cognitive Psychophysiology Lab
> School of Psychology
> The University of Auckland
>
> _______________________________________________
> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> To unsubscribe, send an empty email to eeglablist-unsubscribe at sccn.uc
> sd.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/20171008/26ac6e2f/attachment-0001.html>


More information about the eeglablist mailing list