[Eeglablist] Unequal group size and STUDY use

[Johan]

Dear all,

I'm currently processing about a hundred datasets containing 3 groups
of unequal sizes, and have run today at the same problem. I figured
out that the groups needed to be equal, otherwise the erpdata will
contain NaN's. Glad to see other people already ran into this problem.

The new erpplot std function works!

Regards,

Johan

Johan van der Meer, PhD Student
Dept. Clinical Neurophysiology, room D2-139
Academic Medical Centre
Meibergdreef 9
1105 AZ Amsterdam
Tel: 0031205662965
email: j.n.vandermeer at amc.uva.nl









On Tue, Dec 2, 2008 at 10:32 AM, arno delorme<arno at ucsd.edu> wrote:
> Dear Hamish,
> A related problem was that whenever you have channels without coordinate,
> the interpolation function was not functioning. I have fixed this problem
> and I will send you off the eeglablist some functions to test that.
> In general, I would strongly recommend to use the current version over the
> 6.01 "stable" version of EEGLAB which is more than 1 year and a half old.
> Best,
> Arno
> On 2 déc. 08, at 06:36, Hamish INNES-BROWN wrote:
>
> Interesting, it's working really well for me - I'm very glad!
>
> I did have to ensure that the interpolation was done correctly though.  For
> me this meant using the stable v6.01 to do JUST the interpolation, saving
> the STUDY, and then going back to the Oct08 version to pre-calculate my
> channel measures.  Then the plotting of channel measures with unequal group
> sizes (and I also have two conditions) works fine as far as I can see.
>
> I never did work out why the interpolation didn't work correctly in the
> newer versions.  I tried many different parameters using std_interp,
> eeg_interp as well as the menu system, and always ended up with an error
> (griddata error previously reported).
>
> Anyway that's what worked for me - do interpolation in the stable version of
> eeglab (v6.01), then calculate channel measures in the oct08 version, and
> then plotting will work.  Probably there is an easier way :)
>
> It looks like what you did should have worked though, are you definitely
> using the edited version of std_erpplot.m?  I have a few different eeglab
> directories on my PC and it's easy to start up from the wrong one..
>
>
> Hamish Innes-Brown
> Senior Research Assistant
>
> The Bionic Ear Institute
> c/o St Vincents Hospital
> 6th Floor Daly Wing,
> 35 Victoria Pde, Fitzroy Vic 3065
> Tel: +61 3 9288 3523
> Fax: +61 3 6997 7518
>
>
> ________________________________
> From: Hilla Jakoby [mailto:jakobyhi at yahoo.com]
> Sent: Tuesday, 2 December 2008 2:59 AM
> To: Hamish INNES-BROWN; arno delorme
> Cc: eeglablist at sccn.ucsd.edu; Michael Stevens
> Subject: Re: [Eeglablist] Unequal group size and STUDY use
>
>
> Dear all,
>
> I've tried to use the new std_erpplot.m function with an unequal number of
> subjects of each group (25 versus 29) and the same error message was
> generated when trying to plot their ERPs as before (using the old function).
> I've interpolated all datasets beforehand to 65 electrodes (using
> eeg_interp(EEG, chanlocs,'invdist'). was I right in doing so?
> On equal group sizes STUDY works well.
> Is there another solution for the problem, other changes to be done?
> Many thanks, Hilla
>
>
> Ph.D. student
> Gonda brain reseach center,
> Bar Ilan Universiy, Israel
>
>
>
> --- On Sun, 11/30/08, arno delorme <arno at ucsd.edu> wrote:
>
> From: arno delorme <arno at ucsd.edu>
> Subject: Re: [Eeglablist] Unequal group size and STUDY use
> To: "Hamish INNES-BROWN" <HINNES-BROWN at bionicear.org>
> Cc: eeglablist at sccn.ucsd.edu, "Michael Stevens" <msteven at harthosp.org>
> Date: Sunday, November 30, 2008, 9:40 PM
>
> Dear Michael and Hamish,
>
> thanks for reporting the problem. This issue arises due to the complexity of
> the scheme implemented to handle scalp map plotting in the case of subject
> with
> different number of channels. This scheme was currently enforcing to have
> groups
> of the same size. To fix this problem, I have decided to enforce
> interpolation
> so that all subject will need to have the same number of channels (or at
> least
> subset of channels when plotting scalp maps). Scalp maps where the different
> channels average different numbers of subjects do not make much sense
> anyway.
> There has also been at least two people reporting plotting problems after
> interpolating data, although I have not yet myself been able to reproduce
> the
> problem. I hope the current fix will solve this problem too.
>
> I have reprogrammed the attached std_erpplot.m function so that it can
> handle
> groups of different sizes (although be sure to interpolate missing
> channels).
> Please let me know if this fixes your problems (just replace this function
> in
> the eeglabxxx/functions/studyfunc/ folder) and it will be integrated in the
> EEGLAB monthly release. This message was forwarded to the list because it is
> of
> global interest and also because it depicts a problem that is a major
> current
> limitation for most users. In general though, please use the Bugzilla
> database
> for reporting problems.
>
> Best,
>
> Arno
>
> ps: For interpolation, there is currently a non-documented bug: if you
> interpolate channels when pre-computing ERPs and then go back to the
> pre-compute
> GUI to compute ERSP without interpolation, the function will crash. This is
> because all measure have to be consistent in terms of interpolation. As
> described above, the future function will enforce interpolation (which is
> relatively fast). In case the function crashed for you, erase all measure
> data
> files (*.daterp, *.datersp, *.datspec, *.datitc), and type on the command
> line
> STUDY.changrp = []; then call back the pre-computing interface.
>
> _______________________________________________
> 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
>
>
> _______________________________________________
> 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
>