[pymvpa] null classification performance in the presence of strong univariate signal??

David Soto d.soto.b at gmail.com
Mon Sep 22 21:26:34 UTC 2014


Hi,  when I run the voxelwise group analyses using Jo's code, on the
individual accuracy maps I get nothing,  the highest t value is 0.7 in 3
voxels.....

so the question remains, may it be the case that 8 copes per class are just
insufficient for svm classification?

but still how can it be a null MVPA in the presence of such strong
univariate signal..

cheers
ds

On Fri, Sep 19, 2014 at 2:37 PM, J.A. Etzel <jetzel at artsci.wustl.edu> wrote:

> Yes, that looks like a reasonable single-subject searchlight accuracy map,
> and not the sort of thing that would lead to high significance in all
> voxels in a t-test.
>
> I second Nick's guess that you might have tested against zero instead of
> chance.
>
> Jo
>
>
> On 9/18/2014 5:44 PM, David Soto wrote:
>
>> Thanks Jo and Nick for the advise, the individual acc maps
>> which I registered to MNI prior to t-test look fine to me
>> (see example pic attached, spot with accuracy around 0.7)...there might
>> be something going on in the t-testing done  in FSL, though I cant see
>> what- as should work fine across imaging data types -
>>
>> I will try your R code - thanks!
>>
>> ds
>>
>> On Thu, Sep 18, 2014 at 11:04 PM, J.A. Etzel <jetzel at artsci.wustl.edu
>> <mailto:jetzel at artsci.wustl.edu>> wrote:
>>
>>     I agree with Nick that something might have went wrong with the
>>     t-test. I've never tried one in fsl, either, but usually use R.
>>     Here's a little bit of R code to do a voxelwise t-test:
>>     http://mvpa.blogspot.com/2014/__09/demo-r-code-to-perform-__
>> voxelwise-t-test.html
>>     <http://mvpa.blogspot.com/2014/09/demo-r-code-to-
>> perform-voxelwise-t-test.html>
>>
>>     I assume you looked at the 19 subjects' accuracy maps before trying
>>     the t-test ...
>>
>>     Jo
>>
>
> _______________________________________________
> Pkg-ExpPsy-PyMVPA mailing list
> Pkg-ExpPsy-PyMVPA at lists.alioth.debian.org
> http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/pkg-exppsy-pymvpa
>



-- 
http://www1.imperial.ac.uk/medicine/people/d.soto/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.alioth.debian.org/pipermail/pkg-exppsy-pymvpa/attachments/20140922/19627a14/attachment.html>


More information about the Pkg-ExpPsy-PyMVPA mailing list