[Eeglablist] Error while precomputing STUDY: NaN values in cluster.sets

Makoto Miyakoshi mmiyakoshi at ucsd.edu
Fri Dec 4 10:15:50 PST 2015


Dear Nikola and all,

I learned just yesterday that pop_runica() has an option to 'concatenate
data' when mulitple datasets are selected. However, if you use this option,
the computed ica-related outputs will not be saved correctly. If you
computed your ICA using this approach, be careful.

Makoto

On Wed, Dec 2, 2015 at 10:18 AM, Nikola Vukovic <vukovicnikola at gmail.com>
wrote:

> So I was playing some more with the STUDY files, and I found the solution!
> For some reason, the ICA matrices were not copied correctly from the
> continuous subject-level dataset into the separate epoched sets - despite
> the fact that loading the latter in eeglab shows the ICA as present, and
> even allows the components to be plotted. The solution for me was to
> manually copy the ICA weights from the non-NaN epoched sets to the
> problematic ones. Having done that, it seems to work now - I was able to
> precluster the STUDY components without the error message.
> Hope this helps someone with a similar issue...
> N
>
>
> On Wed, Dec 2, 2015 at 12:33 PM, Sebastian Grissmann <
> sebastian.grissmann at lead.uni-tuebingen.de> wrote:
>
>>
>> Hi there,
>>
>> I had the same issue twice.
>>
>> The only (suboptimal) solution I found
>> was to remove the subjects with NaN values.
>>
>> Best,
>> Sebastian
>>
>>
>>
>> Quoting Nikola Vukovic <vukovicnikola at gmail.com>:
>>
>> Hi Tarik,
>>>
>>> Indeed this seems to be a long standing bug - it has stalled my analysis
>>> for two weeks now... I have found one similar post on Bugzilla a couple
>>> of
>>> years ago, but it was not resolved, and so I posted it again there (
>>> https://sccn.ucsd.edu/bugzilla/show_bug.cgi?id=1822), but it hasn't been
>>> looked at by the admins...
>>>
>>> As for EEGLab not knowing which dataset belongs to which subject, it must
>>> be due to some non-obvious field. At study creation, I tag all datasets
>>> with subject and condition information, like this:
>>> { 'index' 1 'load' '/EEG/RF01/Epoched/RF01_S 1.set' 'subject' 'RF01'
>>> 'condition' 'IAllo' }
>>>
>>> N
>>>
>>> On Tue, Dec 1, 2015 at 5:45 AM, Tarik S Bel-Bahar <
>>> tarikbelbahar at gmail.com>
>>> wrote:
>>>
>>> Greetings,
>>>>
>>>> I've seen this several times, and it's led to a lot of wasted time. I'm
>>>> sure others have come across it before,  perhaps it was already
>>>> mentioned
>>>> it on eeglablist posts, did you check?.
>>>>
>>>> This issue sounds like a bug, if so it should be reported to the EEGLAB
>>>> bugzilla.
>>>>
>>>> I believe that the problem is that eeglab is looking for exactly
>>>> matching
>>>> ICs across datasets that it considers to be coming from one ID or
>>>> subject.
>>>> I would also check if there is not some special information (such as ID
>>>> or
>>>> group) that is being read incorrectly and affecting how STUDY thinks
>>>> about
>>>> organizing the ICs.
>>>>
>>>> Try uploading a small study of short datasets with the ICs, which
>>>> replicates the problem when STUDY is set up.
>>>>
>>>>
>>>>
>>>>
>>>> On Wed, Nov 25, 2015 at 10:41 AM, Akshay Maggu <akshaymaggu86 at gmail.com
>>>> >
>>>> wrote:
>>>>
>>>> Hi Nikola,
>>>>>
>>>>> I am facing the same issue. All pre-processing steps carefully followed
>>>>> and ICAs done for all subjects. But still this 'Na' error (
>>>>> https://www.dropbox.com/s/sw6xvujbalidg24/error_preclustering.JPG?dl=0
>>>>> )
>>>>> pops up saying that some of the datasets have ICA decompositions
>>>>> missing.
>>>>>
>>>>> Let's see what experts on the list have to say about it.
>>>>>
>>>>> Akshay
>>>>>
>>>>> On Tue, Nov 24, 2015 at 7:33 AM, Nikola Vukovic <
>>>>> vukovicnikola at gmail.com>
>>>>> wrote:
>>>>>
>>>>> Dear EEGLAB list
>>>>>>
>>>>>> I would be very grateful is someone could help me out with an error
>>>>>> I've
>>>>>> been getting. When I try to precluster components in a given STUDY (it
>>>>>> happens for several study files), I get the following error:
>>>>>>
>>>>>> FOR CLUSTERING, YOU MAY ONLY USE DIPOLE OR SCALP MAP CLUSTERING. This
>>>>>> is
>>>>>> because some datasets do not have ICA pairs. Look for NaN values in
>>>>>> STUDY.cluster(1).sets which indicate missing datasets. Each column in
>>>>>> this
>>>>>> array indicate datasets with common ICA decompositions.
>>>>>>
>>>>>> Looking inside the cluster(1).sets, I can indeed find some NaN
>>>>>> fields, suggesting not all files have associated ICA decomposition data.
>>>>>> This, however, is puzzling because I made sure to do ICA on each subject's
>>>>>> dataset *before* epoching, meaning that all sets for individual conditions
>>>>>> share the same weights for a subject. After getting this error I even tried
>>>>>> to copy the weights (EEG.icawinv + EEG.icasphere + EEG.icaweights +
>>>>>> EEG.icachansind) from the unepoched set to all the epoched ones.
>>>>>>
>>>>>> Did anyone have this error too, and how could it be solved?
>>>>>>
>>>>>> Thanks for any help,
>>>>>>
>>>>>> Nikola
>>>>>>
>>>>>>
>>>>>> _______________________________________________
>>>>>> 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
>>>>>>
>>>>>>
>>>>>
>>>>>
>>>>> --
>>>>> Akshay Raj Maggu,
>>>>> PhD Candidate,
>>>>> Laboratory for Language, Learning and the Brain,
>>>>> Department of Linguistics and Modern Languages,
>>>>> The Chinese University of Hong Kong,
>>>>> Hong Kong.
>>>>> Webpage: http://www.cuhk.edu.hk/lin/new/people/students/maggu/
>>>>> *Save trees.. Save life! Please don't print this email unless you
>>>>> really
>>>>> need to.....*
>>>>>
>>>>> _______________________________________________
>>>>> 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
>>>>>
>>>>>
>>>>
>>>>
>>
>>
>> --
>> Sebastian Grissmann (Mag. rer. nat)
>> PhD student
>>
>> LEAD Graduate School
>> University of Tübingen
>> Gartenstrasse 29 (1st floor)
>> 72074 Tübingen
>> Germany
>> Phone  +49 7071 29-73579
>>
>> www.lead.uni-tuebingen.de
>>
>>
>> _______________________________________________
>> 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
>



-- 
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/20151204/0ec8a3bf/attachment-0001.html>


More information about the eeglablist mailing list