[Eeglablist] ICA error

[Makoto Miyakoshi]

Dear Erika,

Hmm interesting.
Your procedure seems fine. Maybe it's about process detail by EEGLAB.
How do you recover the original reference channel? Did you do it through
EEGLAB GUI, or via command line without using GUI/EEGLAB functions?

Makoto



> I have 63 channels plus the reference.  After re-referencing to the
average reference and adding the reference channel back to the data there
are 64 channels, the reference channel is channel 64.  I am selecting
channels 1:63 for ICA.  This is when I get the error.  If I select channels
1:64 for ICA it runs fine.

On Wed, Jan 3, 2018 at 11:38 AM, Erika Nyhus <enyhus at bowdoin.edu> wrote:

> Hi Makoto,
>
> I have 63 channels plus the reference.  After re-referencing to the
> average reference and adding the reference channel back to the data there
> are 64 channels, the reference channel is channel 64.  I am selecting
> channels 1:63 for ICA.  This is when I get the error.  If I select channels
> 1:64 for ICA it runs fine.
>
> Erika
>
> On Wed, Jan 3, 2018 at 2:28 PM, Makoto Miyakoshi <mmiyakoshi at ucsd.edu>
> wrote:
>
>> Dear Erika,
>>
>> Your story confused me.
>>
>> > I am selecting 63 channels and excluding the reference channel for
>> ICA.  All the datasets are from the same subject and have been preprocessed
>> in the same way.
>>
>> So you ran ICA on this 63 channels. Hence your EEG.icasphere has 63x63. I
>> guess your EEG.icachansind is also 1x63.
>>
>> > I tried running ICA with 64 channels selected and it runs fine.
>>
>> Where did you get this 64 channel data, and why do you need to run
>> different ICA on this (probably the original, non-channel-removed) data
>> separately?
>>
>> It's like reading a detective story to me.
>>
>> Makoto
>>
>>
>> On Wed, Jan 3, 2018 at 7:21 AM, Erika Nyhus <enyhus at bowdoin.edu> wrote:
>>
>>> Hi Tarik and Makoto,
>>>
>>> All datasets were re-referenced to the average reference and the
>>> reference channel was added back to the data so there are 64 channels and
>>> no bad channels.  I am selecting 63 channels and excluding the reference
>>> channel for ICA.  All the datasets are from the same subject and have been
>>> preprocessed in the same way.  I tried running ICA with 64 channels
>>> selected and it runs fine.  I tried running ICA with the tutorial dataset
>>> and selecting 31 channels and get the same error as in my datasets.
>>>
>>> Erika
>>>
>>> On Tue, Dec 26, 2017 at 6:22 PM, Makoto Miyakoshi <mmiyakoshi at ucsd.edu>
>>> wrote:
>>>
>>>> Dear Erika,
>>>>
>>>> If EEG.icasphere has 63 x 63 columns, it means ICA was computed on 63
>>>> channels. How many channels did you have when you run ICA? Was the data
>>>> full-ranked (no bridging across channels, no standard re-referencing before
>>>> ICA)?
>>>>
>>>> Makoto
>>>>
>>>> On Wed, Dec 20, 2017 at 8:28 AM, Erika Nyhus <enyhus at bowdoin.edu>
>>>> wrote:
>>>>
>>>>> I am getting the following error when I run ICA on a concatenated
>>>>> datasets and selecting 63 of 64 channels.   When I run ICA on any of
>>>>> the datasets individually I do not get an error.  I was wondering if anyone
>>>>> has run in to this problem and how they were able to solve it.
>>>>>
>>>>> eeg_checkset error: number of elements in 'icachansind' (64)
>>>>> does not match the number of columns in the sphere array (63)
>>>>>
>>>>> --
>>>>> Erika Nyhus, Ph.D.
>>>>> Department of Psychology and Program in Neuroscience
>>>>> 6900 College Station
>>>>> Bowdoin College
>>>>> Brunswick, ME 04011
>>>>>
>>>>>
>>>>
>>>>
>>>> --
>>>> Makoto Miyakoshi
>>>> Swartz Center for Computational Neuroscience
>>>> Institute for Neural Computation, University of California San Diego
>>>>
>>>
>>>
>>>
>>> --
>>> Erika Nyhus, Ph.D.
>>> Department of Psychology and Program in Neuroscience
>>> 6900 College Station
>>> Bowdoin College
>>> Brunswick, ME 04011
>>>
>>>
>>
>>
>> --
>> Makoto Miyakoshi
>> Swartz Center for Computational Neuroscience
>> Institute for Neural Computation, University of California San Diego
>>
>
>
>
> --
> Erika Nyhus, Ph.D.
> Department of Psychology and Program in Neuroscience
> 6900 College Station
> Bowdoin College
> Brunswick, ME 04011
>
>


-- 
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/20180108/0954ff90/attachment.html>