[Eeglablist] Question regarding ICA

[Tarik S Bel-Bahar]

My apologies. Sorry to have not been more clear.
It is my (simple) understanding that ICA *does not care *
whether the data is epoched or continuous.
In other words, to ICA, the data looks the same whether it is epoched or
continuous.
ICA is basically trying to find spatial patterns/filters that best explain
the time points you give it.

That being said, there are multiple rules you need to learn about and keep
in mind
before "feeding" data to ICA. The data needs to contain enough time points,
be clean enough, and contain only data from the actual task
(that is, no periods where "nothing was going on" between experimental
blocks).

You may want to do some reading on ICA/EEG fundamentals.
For example, search on Google Scholar for these two items:

1. ERP features and EEG dynamics: an ICA perspective
S Makeig, J Onton - Oxford Handbook of Event-Related Potential …, 2011 -

2. Independent EEG sources are dipolar
A Delorme, J Palmer, J Onton, R Oostenveld, S Makeig - PloS one, 2012 -
dx.plos.org

*Please also take the time to search the eeglablist archives and*
*inform yourself of past conversations. Search on the "keywords" *
*we have been using, such as epoched, continuous, ICA....*
For example, I searched on Google for "ICA continuous or epoched"
and found a few eeglablist conversations, and other pages that
might help you develop a deeper view of what is going on,
including the following from the EEGLAB wiki,
which should be read from start to finish by any new eeglab users.
*
*
* Chapter 09: Decomposing Data Using ICA -
SCCN<http://sccn.ucsd.edu/wiki/Chapter_09:_Decomposing_Data_Using_ICA>
sccn.ucsd.edu/wiki/Chapter_09:_Decomposing_Data_Using_ICA
*

The advice from Stephen and others is also important to keep in mind.













On Wed, Sep 26, 2012 at 1:04 AM, Davide Baldo <davidebaldo84 at gmail.com>
wrote:
>
> Thanks you all for your answer.
>
> Now I am wondering what does that mean tha ICA is agnostic about
continuous/discontinuous data. Could you please explain that concept a bit?
I am very interested in the topic.
>
> Thanks again!
>
> Davide.
>
>
> On Wed, Sep 26, 2012 at 1:17 AM, Tarik S Bel-Bahar <
tarikbelbahar at gmail.com> wrote:
>>
>> yes, periods that are not of cognitive interest should be removed,
>> and ICA is generally agnostic about
>> whether the data is epoched or continuous.
>> yes, there have been various conversations on this topic in the list.
>>
>> On Fri, Sep 21, 2012 at 4:29 AM, Davide Baldo <davidebaldo84 at gmail.com>
wrote:
>>>
>>> Dear all,
>>>
>>> I was wondering about the following problem:
>>>
>>> Assume that your experiment has 2 blocks each with 30 Trials (just a
random numbers). The experiment is divided into 2 blocks in order to give a
break to the subject at the end of the first block of 30 trials.
>>> The EEG data during the break will probably be full of any kinds of
artifacts.
>>>
>>> My question is the following: Regarding ICA, what do you do with the
EEG data recording during the break time? I mean, do you use the complete
EEG signal to run ICA (including the data recorded during the break time)
or do you remove the noise data recording during the break?
>>>
>>> The point is: if you do not remove that data, then you have a
continuous signal as input to ICA.
>>>                    If you do remove that data, you remove a lot of
noise, but you add a discontinuity in the data you use to run ICA (because
you must cut the data recorded during the break and join together the
remaining data)
>>>
>>> Thus...is it better to remove that part of the data or not?
>>>
>>> Does my question make sense? :)
>>>
>>> Thanks in advance for your kind help,
>>>
>>>
>>> Davide.
>>>
>>> _______________________________________________
>>> 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/20120926/1074cdcb/attachment.html>