[Eeglablist] Measure projection

[Makoto Miyakoshi]

Dear Rachel,

When you choose the domain that survived MP statistics, you are already
choosing the one to which limited number of subjects contribute (Nima,
please correct me if I'm wrong). So you don't need to exclude
non-contributors yourself.
MP is the filter to find maximum similarity in both dipole locations and
projected measures.

Makoto


On Tue, Aug 12, 2014 at 3:39 AM, Cooper, Rachel <rcoopea at essex.ac.uk> wrote:

>  Hi Makoto,
>
> Thanks for your reply. More participants were included when using MP than
> clustering (plus there are beautiful plots!). I'm using MP to choose which
> ICs and their activations to analyse from each participant. Do you think it
> is best to exclude participants who didn't contribute any ICs to the MP
> analysis? I could try to find the most similar looking IC from each
> excluded participant but this seems inaccurate and would probably add lots
> of error.
>
> Best
> Rachel
>  ------------------------------
> *From:* Makoto Miyakoshi [mmiyakoshi at ucsd.edu]
> *Sent:* 11 August 2014 19:53
> *To:* Cooper, Rachel
> *Cc:* eeglablist at sccn.ucsd.edu; Nima Bigdely Shamlo
> *Subject:* Re: [Eeglablist] Measure projection
>
>   Dear Rachel,
>
>  > Now that I have run measure projection however, I've found that the
> problem persists.
>
>  If I understand it correctly, MP does not solve the problem completely,
> but it reduces it. If you smooth the data with 3-D Gaussian kernal (default
> with 8 mm, but try 20 mm also) you have *more chance* to overlap more
> subject's ICs in a given 'domain' than not.
>
>  I heard from Tim Mullen that he would release Bayesian Hierachical
> whatever to solve the 'missing data problem' in this approach. We will
> start the test phase soon.
>
>  Makoto
>
> On Wed, Jul 30, 2014 at 4:01 PM, Cooper, Rachel <rcoopea at essex.ac.uk>
> wrote:
>
>>  Hi everyone,
>>
>> Following advice given from members of this list (thank you) I used the
>> measure projection add-on to try to find similar ICs across my
>> participants/conditions. Using measure projection was recommended as a
>> solution to the problem I had when clustering. The problem with clustering
>> was that ICs from some participants didn't appear in a cluster and some
>> participants contributed multiple ICs to a cluster. Now that I have run
>> measure projection however, I've found that the problem persists. Could
>> this be due to a mistake in running the MP analysis? What should I do with
>> the participants who's ICs did not contribute to a domain?
>>
>> Many thanks
>> Rachel
>>
>>
>>
>>  Rachel Cooper
>> PhD researcher
>> Department of Psychology,
>> University of Essex,
>> Wivenhoe Park,
>> Colchester,
>> Essex,
>> CO4 3SQ
>>
>> _______________________________________________
>> 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
>>
>
>
>
>  --
> 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/20140812/f2bc5585/attachment.html>