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

David Soto d.soto.b at gmail.com
Thu Sep 18 13:36:34 UTC 2014

Hi Nick,

I have now carried individual whole-brain searchlight with SVM and
crossvalidation (using the 8 betas in condition A and B from each of the 8
runs as input, no Zscoring as each chunk only has 2 samples, 1 for A and 1
for B).

I reiterate I did this separately for each of the 19 subjects.

 I then aimed to carry out a group analyses using the individual accuracy
maps. to do this I merged the 19 nifti accuracy maps into a 4D file and run
a one-sample t-test in FSL using randomise -i searchpred -o
searchpredOneSampT -1 -v 5 -T.

Weirdly the output gives a group map with all brain voxels over p<0.001 !?
which cannot be right...

Is there any advise please on performing the group level analyses on
individual searchlight accuracy maps?


On Fri, Sep 12, 2014 at 12:10 PM, Nick Oosterhof <
nikolaas.oosterhof at unitn.it> wrote:

> On Sep 11, 2014, at 9:34 PM, David Soto <d.soto.b at gmail.com> wrote:
> > now my question is - would not zscoring solve the issue you mentioned?
> Quite possibly so, but the only way to know is to try.
> >
> > I realise is not possible to zscore by chunks because these are the
> subjects
> > and only 2 samples....but I tried other zscoring schemes (i.e. by task
> condition
> > A vs. B or globally across the 38 samples) and still get null MVPA
> performance
> That's not impossible, but still a bit curious.
> >
> > would there be any advantage in trying the tstats or zstats instead of
> the betas
> > as the former are already zscored?
> It could, with t-stats there is usually less of an issue with outlier
> values.
> _______________________________________________
> 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

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.alioth.debian.org/pipermail/pkg-exppsy-pymvpa/attachments/20140918/9df169b1/attachment.html>

More information about the Pkg-ExpPsy-PyMVPA mailing list