[Eeglablist] Similar ICA components
jumana.ahmad at kcl.ac.uk
Wed Oct 11 10:30:52 PDT 2017
I would not remove the bogus IC. Average mastoid shouldn't reduce your rank.
I reference to the average, and include the original reference (effectively a flag channel) in that reference calculation. I then remove the original reference after the average reference is computed.
If I interpolate electrodes (which I would do to maintain a consistent number of electrodes within the average reference between subjects), then I will use PCA to reduce the dimensionality down to the data rank (number of electrodes not interpolated).
On 11 Oct 2017, at 18:22, mahsa shalchy <mahsa.shalchy at gmail.com<mailto:mahsa.shalchy at gmail.com>> wrote:
When I use average of all channels as my reference I am losing one unit of my rank compared to the case, which I use average mastoid as reference. So, I guess somehow using the average of all electrodes is introducing a kind of dependence between my data matrix rows (channels)! So, I guess I should either use pca to discard the bogus IC or use average mastoid as my reference. My question is which method is preferred? Should I maybe go with the average reference and just discard the bogus IC manually in Ic rejection step?
In addition, when I change the reference many of my ICs change (please see here:
Is this suppose to happen?
On Wed, Oct 11, 2017 at 9:51 AM, Ahmad, Jumana <jumana.ahmad at kcl.ac.uk<mailto:jumana.ahmad at kcl.ac.uk>> wrote:
Can you calculate your data rank?
It should equal the number of non interpolated electrodes put into the ICA. Did you average reference first? Maybe you average referenced, then ran ICA..... in which case your rank would be out by 1, and you'd get a bogus component. If this happens then I normally see the second component as the inverse of the first.
Let me know if any of this is the case, and I can help further if required.
On 11 Oct 2017, at 17:39, mahsa shalchy <mahsa.shalchy at gmail.com<mailto:mahsa.shalchy at gmail.com>> wrote:
I collected some data using Biosemi system (32 channels) and after doing pre-processing steps (resampling, filtering, re-referencing), I epoched the data and applied ICA.
My problem is that, I'm getting two similar components in my ICA results (IC3 and IC4).
I uploaded related images here:
Could you please help me with this?
Eeglablist page: https://emea01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fsccn.ucsd.edu%2Feeglab%2Feeglabmail.html&data=01%7C01%7Cjumana.ahmad%40kcl.ac.uk%7C4ae8a136541542300a9608d510c6a063%7C8370cf1416f34c16b83c724071654356%7C0&sdata=GrPXKHOlQY8pe2%2FNtxV4ioc8FvHN2zz9phb%2BVurt1As%3D&reserved=0
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>
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the eeglablist