[Eeglablist] baseline for short trial durations

Sat Jan 21 23:24:34 PST 2017

Dear Jumana,

> Do you reduce dimensional with PCA first?

Yes, this PCA is to reduce the data dimension for ICA.

> If we want to avoid PCA how should one run AMICA?

You can just skip this PCA stage and just run AMICA/infomax.

> I do not want to add non-linearity to my data. I only want to use ICA to
remove eye blinks from my data.

Does PCA dimension reduction introduce non-linearity into data? Can you
point me to the source (paper, article, etc). I have never heard of it.

> Also, how should one ensure they are not rank deficient after
interpolating.

After interpolation, the data are rank deficient because the interpolated
channels are weighted sum of other channels (i.e. carries no unique
information).

> I will reduce one channel after average referencing, but do I need
further reductions after interpolation?

Yes, if you interpolate 10 channels, the data rank is reduced by 10.
In my new code, it uses the data rank of smaller values of either estimated
rank or number of channels rejected by clean_rawdata(). I found this part
is a little bit tricky, and for unknown reason this rank estimation can go
wrong from time to time.

>From UA49 to Newark,

Makoto

On Fri, Jan 13, 2017 at 10:02 AM, Ahmad, Jumana <jumana.ahmad at kcl.ac.uk>
wrote:

> Dear Makoto,
>
> I have thought about it some more and I will re-do my analysis using your
> pipeline for ICA. Please can you clarify a few things:
>
>
> In your code:
>
>
>  % Step 10: Run AMICA using calculated data rank with 'pcakeep' option
>     if isfield(EEG.etc, 'clean_channel_mask')
>         dataRank = min([rank(double(EEG.data')) sum(EEG.etc.clean_channel_mask)]);
>     else
>         dataRank = min(rank(double(EEG.data')));
>     end
>     runamica15(EEG.data, 'num_chans', EEG.nbchan,...
>         'outdir', ['/data/projects/example/amicaResults/' dataName],...
>         'pcakeep', dataRank, 'num_models', 1,...
>         'do_reject', 1, 'numrej', 15, 'rejsig', 3, 'rejint', 1);
>     EEG.etc.amica  = loadmodout15(['/data/projects/example/amicaResults/' dataName]);
>     EEG.etc.amica.S = EEG.etc.amica.S(1:EEG.etc.amica.num_pcs, :); % Weirdly, I saw size(S,1) be larger than rank. This process does not hurt anyway.
>     EEG.icaweights = EEG.etc.amica.W;
>     EEG.icasphere  = EEG.etc.amica.S;
>     EEG = eeg_checkset(EEG, 'ica');
>
> Do you reduce dimensional with PCA first? If we want to avoid PCA how
> should one run AMICA? I do not want to add non-linearity to my data. I only
> want to use ICA to remove eye blinks from my data.
>
>
> Also, how should one ensure they are not rank deficient after
> interpolating. I will reduce one channel after average referencing, but do
> I need further reductions after interpolation?
>
> Sorry for so many emails about this. Thank you.
>
> Best wishes,
>
> Jumana
>
>
> *------------------------------------------*
> *Jumana Ahmad*
> Post-Doctoral Research Worker in Cognitive Neuroscience
> *EU-AIMS Longitudinal European Autism Project (LEAP) & SynaG Study*
> Room M1.26.Department of Forensic and Neurodevelopmental Sciences (PO 23)
> | Institute of Psychiatry, Psychology & Neuroscience | King’s College
> London | 16 De Crespigny Park | London SE5 8AF
>
> *Phone:* 0207 848 5359| *Email:* jumana.ahmad at kcl.ac.uk
> <antonia.sanjose at kcl.ac.uk> | *Website:* www.eu-aims.eu | *Facebook:*
> www.facebook.com/euaims
>
> ------------------------------
> *From:* Makoto Miyakoshi <mmiyakoshi at ucsd.edu>
> *Sent:* 13 January 2017 07:43:32
> *To:* Ahmad, Jumana
> *Cc:* eeglablist at sccn.ucsd.edu
> *Subject:* Re: [Eeglablist] baseline for short trial durations
>
> Dear Jumana,
>
> > Is this long enough for ITC.
>
> No. See this timely update of my wiki page.
>
> https://sccn.ucsd.edu/wiki/Makoto%27s_preprocessing_
> pipeline#A_tip_to_compute_time-frequency_transform_.28i.
> e._ERSP_.26_ITC.29_.2801.2F11.2F2017_updated.29
> <https://emea01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fsccn.ucsd.edu%2Fwiki%2FMakoto%2527s_preprocessing_pipeline%23A_tip_to_compute_time-frequency_transform_.28i.e._ERSP_.26_ITC.29_.2801.2F11.2F2017_updated.29&data=01%7C01%7Cjumana.ahmad%40kcl.ac.uk%7Cdcb3d49543b1411bf16908d43b7736c8%7C8370cf1416f34c16b83c724071654356%7C0&sdata=UYJEIoWzBT1nzwmNoS%2FQzMbBqbS1%2FPuTM%2FqqZeVSNdE%3D&reserved=0>
>
> Therefore, you still want to use -1 to 2 second even if it create massive
> overlap. Your data will be 6 times heavier, and it's EEGLAB's pipeline
> issue: if you can compute time-frequency transform first then epoch, it
> does not occur. Hoewver, except for inefficiency. you can still perform a
> correct computation, which is good.
>
> Makoto
>
>
>
> On Wed, Jan 4, 2017 at 4:03 AM, Ahmad, Jumana <jumana.ahmad at kcl.ac.uk>
> wrote:
>
>> Dear all,
>>
>>
>>
>> I am analysing data from a MMN oddball experiment with:
>>
>>
>>
>> Standards, 1000Hz 50ms tone; and deviant, 1000Hz, 100ms
>>
>>
>>
>> The SOA is a random 500-600ms, so includes the tones (i.e. doesn't get
>> longer when the duration is longer).
>>
>>
>>
>> I want to calculate ERPs and ITC. What duration would you recommend for a
>> baseline?
>>
>>
>>
>> I am trying to avoid subtracting as much experimental related activity as
>> possible, and as I am focusing on the MMN which occurs fairly early (let’s
>> say before 300ms), I wondered if I should subtract a -100ms baseline. Is
>> this long enough for ITC. Sampling rate is 1000Hz, in most of data (I have
>> 600 datasets, and 10% used 256Hz which of course I have used different
>> filter lengths for etc.). Ideally I would look at the P3 but my trials may
>> not be long enough.
>>
>>
>>
>> Thank you,
>>
>> Jumana
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>> *------------------------------------------*
>>
>> *Jumana Ahmad*
>>
>> Post-Doctoral Research Worker in Cognitive Neuroscience
>>
>> Room M1.09. Department of Forensic and Neurodevelopmental Sciences (PO
>> 23) | Institute of Psychiatry, Psychology & Neuroscience | King’s College
>> London | 16 De Crespigny Park | London SE5 8AF
>>
>>
>>
>> ************************************************************
>> ***************************
>>
>> ************************************************************
>> ***************************
>>
>> *We are currently looking for volunteers with mild intellectual
>> disability to be part of our exciting and world-leading European project
>> into brain development and social behaviour. Please, do get in touch if you
>> know of anyone who may be interested in taking part. *
>>
>> ************************************************************
>> ***************************
>>
>> ************************************************************
>> ***************************
>>
>>
>>
>> _______________________________________________
>> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
>> <https://emea01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fsccn.ucsd.edu%2Feeglab%2Feeglabmail.html&data=01%7C01%7Cjumana.ahmad%40kcl.ac.uk%7Cdcb3d49543b1411bf16908d43b7736c8%7C8370cf1416f34c16b83c724071654356%7C0&sdata=Zdms5ytKQ5G3P2DnQ4kGXZVEDgmB9owLqi%2FvKjFHHHQ%3D&reserved=0>
>> To unsubscribe, send an empty email to eeglablist-unsubscribe at sccn.uc
>> sd.edu
>> For digest mode, send an email with the subject "set digest mime" to
>> eeglablist-request at sccn.ucsd.edu
>>
>
>
>
> --
> Makoto Miyakoshi
> Swartz Center for Computational Neuroscience
> Institute for Neural Computation, University of California San Diego
>

-- 
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/20170121/b2e125b5/attachment.html>