[Eeglablist] Two ICA decompositions from one dataset- based on stage of working memory
irezazadeh at ucdavis.edu
Fri Jun 27 22:24:39 PDT 2014
Here is what I can think as an analogy :
Suppose we have 10 guys and they all talked together in English (from time T1 to T2) and German (from T2 to T3) and we picked their voices by 10 microphones . So if we apply ICA then we will have 10 independent components (voices) ; Ideally, each of the components is from one guy voice from T1 to T3. But ICA is not helpful to distinguish/separate the English and German parts of the conversation from each other.
So here is my thought: if you are interested to study within subject(group) conditions- in the example above the difference btw English and German- it would be useful to use ICA separately on different conditions.
Iman M.Rezazadeh, Ph.D. , M.Sc., B.Sc.
Research Associate II
UCLA David Geffen School of Medicine
Semel Institute for Neuroscience and Human Behavior
760 Westwood Plaza, Ste 47-448
Los Angeles, CA 90095
From: eeglablist-bounces at sccn.ucsd.edu [mailto:eeglablist-bounces at sccn.ucsd.edu] On Behalf Of Makoto Miyakoshi
Sent: Friday, June 27, 2014 10:00 PM
To: Kathleen Van Benthem
Cc: EEGLAB List
Subject: Re: [Eeglablist] Two ICA decompositions from one dataset- based on stage of working memory
> Has anyone tried to compare cluster locations between different conditions within the same experimental session?
That's the other path which is valid but seldom tried. The technical concern is how to make a reasonable comparisons between conditions with non-consistent ICs. However, if a subject participates to the experiment in separate days, the situation in data analysis is more or less the same, so it's definitely not something prohibiting.
On the different note Kathleen, I updated the trimOutlier today and I addressed your request to display name of channels rejected. I also added major improvement in code so that it is now dramatically faster than before (or, well, the former version was too damn...)
On Thu, Jun 26, 2014 at 3:53 PM, Kathleen Van Benthem <kathy_vanbenthem at carleton.ca <mailto:kathy_vanbenthem at carleton.ca> > wrote:
Has anyone tried to compare cluster locations between different conditions within the same experimental session?
We are interested in locating dipoles associated specifically with encoding versus maintenance during a working memory task, and are using EEGLAB to conduct ICA and then PCA to examine dipoles.
Our first method did not give us much success--originally, we tried running ICA on the entire dataset (including both encoding and maintenance conditions), and epoched after ICA was run, so that we ended up with separate files for the two conditions, each with the same ICA weights. When we ran STUDIES for each condition separately, we found the resulting clusters were the same between encoding and maintenance conditions, even though we expected differences to exist.
Because we didn't experience success with this method, we are wondering whether it would be advisable to create two separate files for each participant (from the same session) BEFORE running ICA-- One file would contain EEG data epoched from the first condition (encoding--two seconds per trial), and the second EEG data epoched from the second condition (maintenance--2 seconds per trial). We would then run ICA separately for each file and look to see if there were dipoles from the encoding condition ICA that consistently were not in the maintenance condition ICA results (and vice versa). We have a priori hypotheses of where maintenance activity would be taking place source-wise as compared to where encoding would be taking place. We are wondering if it is advisable to compare dipoles that result from different ICA weights- even when they are collected from the same session and just seconds apart. If there were unique sources of brain activity (dipoles) would it be possible, from this design, to conclude that the unique dipoles were in fact associated with the different stages of the working memory task (all other factors being held constant)?
Thanks for your thoughts on this. We are aware that we need to make sure we have large enough files in order to conduct ICA in the first place.
Kathleen (Kathy) Van Benthem M.H.S., B.Sc.O.T
Carleton University, Institute of Cognitive Science, Ph.D. Candidate,
Cognitive Science, Carleton University
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>
Swartz Center for Computational Neuroscience
Institute for Neural Computation, University of California San Diego
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the eeglablist