[pymvpa] permutation test for unbalanced data

[Yaroslav Halchenko]

> 1.  Unfortunately "lazy" us didn't implement such a dedicated error
>     function yet, BUT  you have that  value as a part of the 

>     In [84]: print cm.stats['mean(TPR)']
>     0.5

>     which is the mean of TPRs per each label, which in turn is what
>     we are interested here

FWIW now it is coming: https://github.com/PyMVPA/PyMVPA/pull/541
and it might better fit your usecase.  Chance distribution should be
more "reliably" centered around chance of 0.5 regardless of the
classifier.  I would be happy to be proven wrong

-- 
Yaroslav O. Halchenko
Center for Open Neuroscience     http://centerforopenneuroscience.org
Dartmouth College, 419 Moore Hall, Hinman Box 6207, Hanover, NH 03755
Phone: +1 (603) 646-9834                       Fax: +1 (603) 646-1419
WWW:   http://www.linkedin.com/in/yarik