[Eeglablist] Reject data (all methods)

Mon Apr 1 18:14:30 PDT 2013

Dear Ida,

> Since I already calculated ICA algorithm on continuous data, does it make
sense to epoch the data now, before second ICA, what would it "do" to
components? Or it is better to do the epoching before first ICA and do it
again?

Once ICA is computed, epoching does not affect the ICA matrices. If your
events of interest is embedded with large amount of data of no interest,
then you'd better epoch first and then run ICA, otherwise ICA will focus on
the noninteresting data.

>At this moment I see also this option - Since I read that the experience
of some researchers is that second ICA does not mean that much better
decomposition - I could remove components (calculated on continuous data)
recognized as artifacts and prune the dataset and see what I have.

If you use STUDY you don't have to remove ICs manually at the first level
(individual subject level). It will be filtered by residual variance and
out of the head criterion, and then clustered into artifacts and EEG
signals.

Makoto

2013/3/30 ida miokovic <ida.miokovic at gmail.com>

> Hello everyone,
>
> thank you Marco, Tarik, Makoto, Arnaud for your suggestions.
>
> I will surely consider using regepochs function. As I understood its
> functioning correctly - it is epoching data set in way that it is possible
> to merge them together later. Since I already calculated ICA algorithm on
> continuous data, does it make sense to epoch the data now, before second
> ICA, what would it "do" to components? Or it is better to do the epoching
> before first ICA and do it again?
>
> At this moment I see also this option - Since I read that the experience
> of some researchers is that second ICA does not mean that much better
> decomposition - I could remove components (calculated on continuous data)
> recognized as artifacts and prune the dataset and see what I have.
>
> I appreciate your opinion.
>
> Thanks
>
> Ida
>
>
> On Thu, Mar 28, 2013 at 10:26 AM, Arnaud Delorme <arno at ucsd.edu> wrote:
>
>> Dear Ida,
>>
>> there is the menu "Tools > Automatic continuous rejection" although this
>> does not use ICA component activities. It would be a good idea to add
>> component to that function as well as channels.
>>
>> Arno
>>
>> On 27 Mar 2013, at 17:53, Makoto Miyakoshi wrote:
>>
>> Dear Ida,
>>
>> >Are there any suggestions how to implement any kind of "bad" parts of
>> the components removal, when having continuous data (before running second
>> ICA)?
>>
>> When you plot raw EEG, you can choose time period (marked as light blue)
>> to register as bad data. You can reject the marked data portion later.
>>
>> If you are interested in automated algorithm for that you may want to
>> check BCILAB.
>>
>> Makoto
>>
>> 2013/3/24 ida miokovic <ida.miokovic at gmail.com>
>>
>>> Dear Makoto, Mikolaj and Marco (and everyone else =)),
>>>
>>> thank you very much for your answers.
>>>
>>> Regarding 1st issue - Yes, I have been informed that there is a bug that
>>> causes this channels locations problem, I have seen that others reported
>>> the same issue, so I didn't reported it myself since it was said that this
>>> will be corrected in the next EEGLAB version. Marco, is it necessary for me
>>> to report this in order to get the developers version? How could I get it?
>>>
>>> Regarding 2nd issue - I didn't realize that mentioned methods are
>>> applicable only on epoched data. Are there any suggestions how to implement
>>> any kind of "bad" parts of the components removal, when having continuous
>>> data (before running second ICA)? Does EEGLAB support that? If not, any
>>> pipeline in MATLAB itself would be appreciated...
>>>
>>> Thank you very much again
>>>
>>> Ida
>>>
>>>
>>> On Fri, Mar 22, 2013 at 4:32 PM, Marco Montalto <montaltomarco at onvol.net
>>> > wrote:
>>>
>>>> Dear Ida,
>>>>
>>>> Regarding Problem no. 1, I was having the same problem with the same
>>>> version of EEGLAB you are using. When I plotted channel locations by
>>>> pressing plot-2D I used to get a plot with all channels squashed in the
>>>> bottom left hand corner of the plot. I reported the problem. Arno
>>>> reproduced the problem and made the necessary changes to the developer's
>>>> version of EEGLAB. I am using the developer's version now and the problem
>>>> is indeed solved. So I would recommend that you use the developer's version
>>>> to overcome this problem.
>>>>
>>>> Regarding Problem no. 2, (corroborating what was said previously
>>>> by Mikołaj Magnuski), Reject data (all methods) is greyed out because you
>>>> have not extracted any epochs. Since you are using continuous data you
>>>> cannot make use of this function, .
>>>>
>>>> Regards,
>>>> Marco
>>>>
>>>>
>>>> 2013/3/17 ida miokovic <ida.miokovic at gmail.com>
>>>>
>>>>> Dear list,
>>>>>
>>>>> I am running ICA decomposition on continuous data. I have several
>>>>> questions regarding this:
>>>>>
>>>>> 1. After applying lowpass filter and removing line noise (using
>>>>> Cleanline plug-in), I import the channel locations of the dataset. I am
>>>>> using EEGLAB v12.0.1.0b which is having issues with channel locations
>>>>> importing, but when I obtain ICA components, scalp maps seem to be plotted
>>>>> correctly, so I suppose that this bug does not influence this step? After
>>>>> that, I do the average reference.
>>>>>
>>>>> 2. Since I plan to run ICA second time, I should remove bad trials of
>>>>> the components (not entire component/s) using Tools --> Reject data using
>>>>> ICA --> Reject data (all methods). After doing all described above, I do
>>>>> not have an option to click on this tab "Reject data (all methods)" (it is
>>>>> grey). It is possibe to click only on "Reject components by map". Does
>>>>> anyone have an idea what could be the reason for this?
>>>>>
>>>>> 3. There is a suggestion that not more than 10% of the dataset should
>>>>> be rejected after applying all the methods I am interested in. Are some of
>>>>> the methods given it this menu (Find abnormal values, Find abnormal
>>>>> trends, Find impropable data, Find abnormal distributions, Find abnormal
>>>>> spectra (slow)) more convenient that others and which? Which way I can
>>>>> calculate what is the percent of the rejected data?
>>>>>
>>>>> 4. Should the mean of the all channels be removed somewhere in this
>>>>> procedure and where is it suggested to be done? As I mentioned, here
>>>>> contiuous data is used, so there are no epochs and no need for baseline
>>>>> removal.
>>>>>
>>>>>
>>>>> Thank you very much for any help.
>>>>>
>>>>> All the best
>>>>>
>>>>> Ida
>>>>>
>>>>>
>>>
>>> _______________________________________________
>>> 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
>> JSPS Postdoctral Fellow for Research Abroad
>> Swartz Center for Computational Neuroscience
>> Institute for Neural Computation, University of California San Diego
>>  _______________________________________________
>> 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/20130401/fb9e4441/attachment.html>