[Eeglablist] Data change when updating eeglab from version 10 to 12

Arnaud Delorme arno at ucsd.edu
Mon Sep 16 04:23:36 PDT 2013


Yes, increasing the number of permutation should make the result more stable.
The default number of permutation is 200. When using FDR correction for multiple comparisons, the default number of permutation is bumped to 4000.
Best,

Arno

On Sep 16, 2013, at 9:43 AM, Ilana Podlipsky <ilana.mlist at gmail.com> wrote:

> Dear Arno,
> 
> Thank you for your reply.
> Would increase of the number of permutations make the result more stable?
> Is there some rule of thumb for a suggested number of permutation for a given data set?
> 
> Thank you
> Ilana
> 
> 
> On Sun, Sep 15, 2013 at 10:23 PM, Arnaud Delorme <arno at ucsd.edu> wrote:
> Dear Llana,
> 
> this is because the permutation statistics use different random permutation every time.
> The result should be similar but not identical.
> Best,
> 
> Arno
> 
> On Sep 12, 2013, at 11:33 AM, Ilana Podlipsky <ilana.mlist at gmail.com> wrote:
> 
>> Dear Nicolas and Arno,
>> I recently ran an ANOVA statistic of ERSP maps across conditions and encountered the same problem of inconsistency of statistics results each time I run the analysis.
>> In short, every time I run the permutation statistics analysis I get different significant electrodes.
>> I encountered this issue in EEGLAB 8,10 and 12 (didn't try 11).
>> I see now that Nicolas reported exactly the same problem.
>> Were you able to solve it?
>>  
>> Thank you for your help,
>> Ilana
>>  
>>  
>> 
>> 
>> On Fri, May 3, 2013 at 7:33 AM, Arnaud Delorme <arno at ucsd.edu> wrote:
>> Dear Nicolas,
>> 
>> your problem is worth investigating. I have checked with the tutorial STUDY, and the results with parametric statistics (EEGLAB) are strictly identical from one call to the next when plotting channel ERSP time-frequency images or channel ERSP scalp topographies. Also since 2010, we have a battery of independent test that check EEGLAB statistics agains the Matlab statistics toolbox. After adding Fieldtrip statistics to the EEGLAB graphic interface  last year, we also implemented the same tests for these new functions (which is how we discovered a bug in Anova Fieldtrip statistics by the way - currently in the process of being fixed). 
>> This said, no program is perfect. Please upload your data to Bugzilla so we can try reproduce the problem and see what is going on.
>> 
>> http://sccn.ucsd.edu/eeglab/bugzilla
>> 
>> If your data is really large (>1Gb), put it on our public FTP server (ask Derrick Lock <dlock at ucsd.edu> for the address as I prefer to write it here for security reasons). Let's also continue this discussion off the list.
>> 
>> Best regards,
>> 
>> Arno
>> 
>> On 2 May 2013, at 09:30, Nicolas Rochet wrote:
>> 
>>> Dear Makoto and Arno,
>>> To sum up: i'm using study analysis using channels (with GUI) to compare two conditions. After further testings, i have similar issues with EEGLAB version 12 and 11.05.5.4b : the stats maps (for condition comparaison in ERSP / ITC analysis) are changing each time i replot whatever stats option i use. Also, while ERP's topographics map are not changing, the significant channels (when comparing the two conditions) are changing each time i replot.
>>> Strangely, those issues only happens with EEGLAB statistic's option, indeed when i choose Fieldtrip stat's option, i don't have this problem: the stats map remains the same after each replotting.
>>> 
>>> cordially,
>>> Nicolas
>>> 
>>> Le 30/04/2013 17:00, Makoto Miyakoshi a écrit :
>>>> Dear Nicolas and Arno,
>>>> 
>>>> > some of these data, the stats are changing, whatever stats option i used (parametric, permutations, with or without holms correction and bonferoni correction). 
>>>> 
>>>> Permutation results may change because this approach is based on random permutation. However parametric test result should not. Could you check it again?
>>>> 
>>>> > ERP's topographic map's ( the significant channels are not the same each time i replot).
>>>> 
>>>> Significant channels? You mean you are running STUDY analysis using channels? If that's the case I'm not experienced with it.
>>>> 
>>>> However, changing topographic maps is familiar with me. However, I want to know exactly what you are doing (using ICs or channels)
>>>> 
>>>> Makoto
>>>> 
>>>> 2013/4/26 Nicolas Rochet <nicolas.rochet at univ-provence.fr>
>>>> Hi Makoto,
>>>> Firstly, I recomputed ERP, ERSP and ITC after updating to version 12. I realised that (when using the study's GUI with version 12) each time i plot some of these                   data, the stats are changing, whatever stats option i used (parametric, permutations, with or without holms correction and bonferoni correction). 
>>>> For example. when i try to compare ERSP maps in two different condition: each time i replot it, the corresponding stat map is changing while the ERSP map is the same. I had the same problem with ITC maps or for ERP's topographic map's ( the significant channels are not the same each time i replot).
>>>> 
>>>> Do you have any idea ? Nobody seems to have reported similar issues ....
>>>> 
>>>> Nicolas
>>>> 
>>>> Le 18/04/2013 03:26, Makoto Miyakoshi a écrit :
>>>>> Dear Nicolas,
>>>>> 
>>>>> I'd be interested in learning how much the difference was. Would you mind explaining it to us?
>>>>> 
>>>>> Makoto
>>>>> 
>>>>> 
>>>>> 2013/4/15 Nicolas Rochet <nicolas.rochet at univ-provence.fr>
>>>>> Dear EEGLAB community,
>>>>> I was previously using eeglab version 10.2.2.4b. After updating EEGLAB
>>>>> to  version 12.0.2.0b (with SVN) the SAME data (ie  ERP, ERSP and ITC
>>>>> grand average) processed with the study protocol are different with the
>>>>> two version
>>>>> Does anybody encountered similar issues ?
>>>>> 
>>>>> Thanks,
>>>>> Nicolas
>>>>> 
>>>>> --
>>>>> Nicolas Rochet, PhD Student
>>>>> Laboratoire de Neurosciences Cognitives, UMR 7291
>>>>> 
>>>>> Université de Provence / CNRS
>>>>> Pôle 3C - Case C
>>>>> 3 place Victor Hugo
>>>>> 13331 Marseille Cedex 03
>>>>> France
>>>>> 
>>>>> tel : (+33) 04 13 55 09 00
>>>>> 
>>>>> _______________________________________________
>>>>> 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
>>>> 
>>>> 
>>>> -- 
>>>> Nicolas Rochet, PhD Student
>>>> Laboratoire de Neurosciences Cognitives, UMR 7291 
>>>> 
>>>> Université de Provence / CNRS
>>>> Pôle 3C - Case C
>>>> 3 place Victor Hugo
>>>> 13331 Marseille Cedex 03
>>>> France
>>>> 
>>>> tel : (+33) 04 13 55 09 00
>>>> 
>>>> _______________________________________________
>>>> 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
>>> 
>>> 
>>> -- 
>>> Nicolas Rochet, PhD Student
>>> Laboratoire de Neurosciences Cognitives, UMR 7291 
>>> 
>>> Université de Provence / CNRS
>>> Pôle 3C - Case C
>>> 3 place Victor Hugo
>>> 13331 Marseille Cedex 03
>>> France
>>> 
>>> tel : (+33) 04 13 55 09 00
>>> _______________________________________________
>>> 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
>> 
> 
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20130916/69f858db/attachment.html>


More information about the eeglablist mailing list