[Eeglablist] Epoching eeg data

Tsatsishvili Valeri valeri.v.tsatsishvili at jyu.fi
Wed Sep 11 00:51:06 PDT 2013


Just wanted to add small side note to the question/answer quoted below:

> The previous messages largely answer my query, but I'd like just to make sure I understand correctly. The results of ICA decomposition should be similar whether the data is epoched or not- is that correct?

> Correct. If you have the same number of datapoint before and after epoching, ICA results should be as identical as the two results from running ICA twice on the same data.


I think generally ICA won't give identical results when run several times even on the same data as it is a stochastic process. Otherwise there would not exist software packages that assess ICA decomposition stability (e.g. ICASSO).



________________________________
From: eeglablist-bounces at sccn.ucsd.edu [eeglablist-bounces at sccn.ucsd.edu] on behalf of Makoto Miyakoshi [mmiyakoshi at ucsd.edu]
Sent: Wednesday, September 11, 2013 4:06 AM
To: Katherine Naish
Cc: eeglablist at sccn.ucsd.edu
Subject: Re: [Eeglablist] Epoching eeg data

Dear Katherine,

> The previous messages largely answer my query, but I'd like just to make sure I understand correctly. The results of ICA decomposition should be similar whether the data is epoched or not- is that correct?

Correct. If you have the same number of datapoint before and after epoching, ICA results should be as identical as the two results from running ICA twice on the same data.

> Obviously the results will be different if the data itself is different, but does ICA essentially treat epoched data as a continuous section?

ICA shuffles all datapoints as a preprocess; thus the concept of time disappears.

> My pre-processing order is the following: resample, filter, epoch, remove baseline, interpolate bad channels and remove bad epochs (based on visual inspection), re-reference to average, and (finally) run ICA to detect artefacts.

It looks ok.

> One thing I wasn't sure about was re-referencing- am i doing that at the right point?

Yes.

Makoto

2013/9/10 Katherine Naish <K.R.Naish at pgr.reading.ac.uk<mailto:K.R.Naish at pgr.reading.ac.uk>>

Dear all,



The previous messages largely answer my query, but I'd like just to make sure I understand correctly. The results of ICA decomposition should be similar whether the data is epoched or not- is that correct? Obviously the results will be different if the data itself is different, but does ICA essentially treat epoched data as a continuous section?



I would also really appreciate it if someone could approve, or otherwise, my pre-processing order. My pre-processing order is the following: resample, filter, epoch, remove baseline, interpolate bad channels and remove bad epochs (based on visual inspection), re-reference to average, and (finally) run ICA to detect artefacts. One thing I wasn't sure about was re-referencing- am i doing that at the right point?



Many thanks in advance,

katherine




________________________________
From: eeglablist-bounces at sccn.ucsd.edu<mailto:eeglablist-bounces at sccn.ucsd.edu> [eeglablist-bounces at sccn.ucsd.edu<mailto:eeglablist-bounces at sccn.ucsd.edu>] on behalf of Mikołaj Magnuski [imponderabilion at gmail.com<mailto:imponderabilion at gmail.com>]
Sent: 06 June 2013 19:13
Cc: eeglablist at sccn.ucsd.edu<mailto:eeglablist at sccn.ucsd.edu>
Subject: Re: [Eeglablist] Epoching eeg data

Dear Karlo,

there is a case when dividing your whole 15 min. into smaller epochs (say 1 or 2 seconds long)
is quite convenient with respect to ICA:
--> cleaning data before ICA is easy if you just have to click bad epochs instead of marking
bad periods in continuous data - although this is just my personal feeling.
Nevertheless, you can also use automatic artifact rejection methods
available in EEGlab this way. These methods require epoched data.
(it is also easy to reconstruct your rejections later if you do them
on epochs - you can just save the indices of windows that you removed [by
coping relevant field in EEG structure after marking bad epochs,
but before rejecting them]).
--> Then, after your first IC decomposition, if you are not satisfied with its results and observe that
some parts of your data may be responsible for poor results - you can remove
epochs containing these data easily and perform ICA again.

So, just as Makoto wrote, ICA does not care whether data are epoched but you may -
because in some cases it is convenient in preprocessing or 'postprocessing' (correcting
ICA decomposition).
If you need any extra help with this - let me know (I have some scripts and functions to do this).

Mikolaj


W dniu środa, 5 czerwca 2013 użytkownik Makoto Miyakoshi napisał:
Dear Karlo,

If you want to epoch your data into 3-min blocks, for example, put event markers every 3 minutes then epoch the data.

ICA does not care temporal continuity of the data since it shuffles them up differently for every iteration as a preprocess. That means, if the total number of datapoints are the same before and after the epoching, the ICA results should be the same.

Makoto


2013/6/3 karlo gonzales <thats_karlo at yahoo.com<https://dbxprd0111.outlook.com/owa/UrlBlockedError.aspx>>
Dear EEGlab experts,

I need you help regarding an eeg data files which was recorded during eye closed state (15 min long).  Thus, there is no event to epoch the data. 1) How do you epoch such data?  2) should we epoch data before running ica (dose more epoch means better results?)
Thanks in advance

_______________________________________________
Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
To unsubscribe, send an empty email to eeglablist-unsubscribe at sccn.ucsd.edu<https://dbxprd0111.outlook.com/owa/UrlBlockedError.aspx>
For digest mode, send an email with the subject "set digest mime" to eeglablist-request at sccn.ucsd.edu<https://dbxprd0111.outlook.com/owa/UrlBlockedError.aspx>



--
Makoto Miyakoshi
Swartz Center for Computational Neuroscience
Institute for Neural Computation, University of California San Diego


--
Pozdrawiam,
Mikołaj Magnuski

_______________________________________________
Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
To unsubscribe, send an empty email to eeglablist-unsubscribe at sccn.ucsd.edu<mailto: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<mailto: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/20130911/cdfac91b/attachment.html>


More information about the eeglablist mailing list