[Eeglablist] [External] Re: Scaling issue when copying ica weights/sphere.

[Fine, Justin Michael]

Hi Makoto:

I had triple checked both of the things you mentioned before writing my original post.  

Regarding #2: I had even run eeg_checkset(EEG, ‘ica’) to be specific about it.  I even tried it with RMS scaling commented out and then back on (original code).  

None of that worked.  I checked the matrices were copied correct as well.  

Note, I even manually rescaled the ica matrices by a factor of 100 or 1000 and that seemed to look better.  

Instead of messing with it much more, I just went ahead and reprocessed the data as a continuous dataset rather than epoching before ICA.  

Justin M. Fine, PhD

> On Jul 18, 2020, at 4:38 PM, Makoto Miyakoshi <mmiyakoshi at ucsd.edu> wrote:
> 
> This message was sent from a non-IU address. Please exercise caution when clicking links or opening attachments from external sources.
> -------
> 
> Dear Justin,
> 
> Off the top of my head, check the following two possible causes
> 
> 1. You made a mistake in copying the weight matrices
> https://sccn.ucsd.edu/wiki/Makoto's_preprocessing_pipeline#Tips_for_copying_ICA_results_to_other_datasets_.2806.2F26.2F2018_updated.29
> 
> 
> 2. It's a rare issue, but you got trapped with the RMS issue. I discussed
> it with Seyed just a couple of weeks ago.
> https://sccn.ucsd.edu/wiki/Makoto's_preprocessing_pipeline#RMS_option_issue_.2807.2F02.2F2020_added.29
> 
> 
> If your problem is different from any of these, write me again to the list.
> 
> Makoto
> 
> 
>> On Tue, Jul 14, 2020 at 2:41 AM Fine, Justin Michael <justfine at iu.edu>
>> wrote:
>> 
>> Hello experts:
>> 
>> I cleaned and ran ica on an epoched dataset. The components looked good
>> all around.
>> 
>> 1. When I went to copy the ica weight and sphere matrices to the same
>> dataset that was continuous, the components activation is now super high.
>> On the order of 50x greater, and it looks poorly.
>> 
>> 2. However , when I copy the weight matrices from an epoched to another
>> epoched dataset, it looks normal.
>> 
>> Any thoughts on where I want wrong?  The scaling is wrong somewhere, as
>> clearly people copy ica weights from epoched to continuous data normally.
>> 
>> Justin M. Fine, PhD
>> _______________________________________________
>> 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
>> 
> _______________________________________________
> 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
>