[Eeglablist] PREP pipeline & boundary events

李想 lxraplikelx at qq.com
Fri Sep 18 16:19:28 PDT 2015


Dear Nima

So I don't really need to reject interpolated channels. 
Thank you Nima.

Li Xiang

> 在 2015年9月19日,00:28,Nima Bigdely Shamlo <nima.bigdely at syntrogi.com> 写道:
> 
> Dear Li,
> 
> It is perfectly OK to use interpolated data as long as you do a PCA to the real rank of the data before the ICA process. To do this first you need to calculate the rank of the data:
> 
> r = rank(double(EEG.data(:,:));
> 
> and then in your ICA options, e.g. in Runica or Binica, provide this rank as the pca dimentions with ...'pca, r.
> 
> -Nima
> 
> 
> 
> 
> 
>> On Thu, Sep 17, 2015 at 7:40 PM, Makoto Miyakoshi <mmiyakoshi at ucsd.edu> wrote:
>> Dear Li,
>> 
>> > But, for the third question, I am still confused that whether it is ok for ica using interpolated data. Could you please explain whether there is difference in ica results between interpolated and uninterpolated data?
>> 
>> Channel interpolated data are not full ranked i.e. some of the channels are just combination of others so do not carry unique information. ICA may be confused if it regards the interpolated data as full-rank... and we confirmed that in some cases it does cause problems, such as 'ghost ICs' which you can find here
>> http://sccn.ucsd.edu/wiki/Makoto's_preprocessing_pipeline#Re-reference_the_data_to_average
>> 
>> Makoto
>> 
>>> On Sat, Sep 12, 2015 at 1:27 AM, lxraplikelx at qq.com <lxraplikelx at qq.com> wrote:
>>> 
>>> 
>>> Dear Nima & Makoto
>>> Thanks for the kind help!
>>> Yeah, I thought it would be no big problem if I use merged dataset before PREP pipeline.
>>> But, for the third question, I am still confused that whether it is ok for ica using interpolated data. Could you please explain whether there is difference in ica results between interpolated and uninterpolated data?
>>> Thank you for your time! 
>>> Li Xiang
>>> 
>>> From: Nima Bigdely Shamlo
>>> Date: 2015-09-12 06:21
>>> To: Makoto Miyakoshi
>>> CC: lxraplikelx at qq.com; eeglablist
>>> Subject: Re: [Eeglablist] PREP pipeline & boundary events
>>> Dear Li
>>> 
>>> (i) You could, especially if the data a few seconds before/after the boundary event is not that important.
>>> (ii) No (at least major) problems.
>>> (iii) No, since each dataset will be interpolated to have the original number of channels (bad channels replaced by interpolated ones). You will have the same number of channels in all blocks. You will only need to do a 'pca' to the rank of the merged data (in double precision) befoe doing the ICA. 
>>> 
>>> -Nima
>>> 
>>> 
>>> 
>>> 
>>>> On Wed, Sep 9, 2015 at 6:34 PM, Makoto Miyakoshi <mmiyakoshi at ucsd.edu> wrote:
>>>> Dear Nima,
>>>> 
>>>> Here is a question for you about the PREP pipeline. Would you mind helping him?
>>>> 
>>>> Makoto
>>>> 
>>>>> On Fri, Sep 4, 2015 at 5:20 AM, lxraplikelx at qq.com <lxraplikelx at qq.com> wrote:
>>>>> Dear eeglablist
>>>>> 
>>>>> I get into an awkward situation using PREP pipeline. The raw datasets I am dealing with are stored separately based on blocks. Firstly, I merged them up for every subject. Then I used PREP pipeline without ignoring boundary event. Since PREP does respect boundary events which represent discontinuities in dataset, PREP skiped all procedures. 
>>>>> 
>>>>> SO, the situation is that 
>>>>> (i)when using merged dataset, whether should I just ignore boundary events or not? 
>>>>> (ii)If I truely ignoring boundary events and run PREP, is there any problem? 
>>>>> (iii)If I do not merge dataset and use datasets of single block, it will very likely lead to inconsistent number of channels in dataset of subject level due to different number of bad channels, and that will not be legitimate in the downstream behavior like ICA. 
>>>>> 
>>>>> What should I do? 
>>>>> 
>>>>> Thank you!
>>>>> 
>>>>> Best!
>>>>> Li Xiang
>>>>> 
>>>>> _______________________________________________
>>>>> 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
>>> 
>>> 
>>> 
>>> 
>>> Sent with MailTrack
>>> 
>> 
>> 
>> 
>> -- 
>> 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/20150919/c155502c/attachment.html>


More information about the eeglablist mailing list