[Eeglablist] Data rank and ICA results

[Makoto Miyakoshi]

Dear Shiva,

> I also have another question, I am going to use ASR to correct
larg artifacts in my continuous signal but I don't want to reject any data
point.
My question is using ASR (with SD=20) lead to eye blink correction most of
the time. Is this procedure good for the proceeding AMICA which I always
use to reject eye blinks or not? which one is more correct for eye blink
rejection?

It is definitely better to use ICA to reject eye blink artifacts.
However, because ICA is a stationary method, it can be easily confused by,
for example, a short-duration high-amplitude artifact that is never
repeated.
ASR is good at removing such a sporadic type of artifact which ICA can
never decompose. It is because ASR is a non-stationary method using a
sliding window.
So ASR is ideal for a preprocessing for ICA. ASR's non-stationary nature
complements ICA well.

Makoto

On Fri, Jul 3, 2020 at 4:12 PM Makoto Miyakoshi <mmiyakoshi at ucsd.edu> wrote:

> Dear Shiva,
>
> Hi this is Makoto. Good to hear from you again! Hope you are doing well.
>
> Assuming your electrodes are not bridged, I think my comment is the same
> as I wrote yesterday for Ivonne. Let me paste it below.
> I recommend you check the smallest value of eig(cov(data))--if it is
> smaller than the heuristically set threshold in runica/amica(which I
> haven't checked myself), it will be cut.
>
> Makoto
>
> %%%%%%%%%%%%%%%%%%%%%%
> Dear Ivonne,
>
> The smallest eigenvalue of 10^-20 is still regarded as 'independent' in
> terms of rank() but for practical application of ICA the algorithm may
> behave as if rank-deficiency is present. I believe I discussed this issue
> with Jason Palmer last time. Isn't the current 'heuristic' minimum
> eigenvalue cutoff something like 10^-8? You should be able to find it in
> runica() function, if I remember correctly.
>
> If you can specify data rank by using the pca option, that would be the
> best to avoid this kind of problems.
>
> So what you described seems correct, except
>
> >  (deficient by 1 due to re-referencing)
>
> A proper re-referencing should not reduce rank. See below.
>
> https://sccn.ucsd.edu/wiki/Makoto's_preprocessing_pipeline#Why_should_we_add_zero-filled_channel_before_average_referencing.3F_.2803.2F04.2F2020_Updated.29
>
>
> Makoto
>
>
> On Thu, Jul 2, 2020 at 3:03 PM shiva khoshnoud <shiva.khoshnoud at gmail.com>
> wrote:
>
>> Hello everyone,
>>
>> I have a question regarding EEG data rank and ICA analysis. I have an EEG
>> set file with 31 channels recorded with Fz electrode as the reference. I
>> did the average re-referencing with adding the initial reference with zero
>> data value to the data by appending one channel  Fz, setting it as
>> reference and adding current reference back to Fz while re-referencing.
>> This procidure gave me a set file with 32 channels including Fz with data
>> rank 32.
>> Then I did ICA with AMICA. In most sets I got 31 components but in some I
>> got 32 components. Considering this fact that my data rank is now 32,
>> shouldn't I get 32 ICs?
>>
>> Thank you so much,
>> Shiva
>>
>>
>>
>>
>>
>>
>> -------------------------------------
>>
>> Shiva Khoshnoud
>>
>> Postdoctoral researcher,
>>
>> Institut für Grenzgebiete der Psychologie und Psychohygiene
>>
>> Wilhelmstr. 3a
>> 79098 Freiburg
>> Germany
>>
>>
>> Tel. +49 761 20721-71
>> Fax. +49 761 20721-91
>> _______________________________________________
>> 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
>
>