[pymvpa] Searchlights and Permutation testing

Yaroslav Halchenko debian at onerussian.com
Wed Jun 17 18:26:43 UTC 2015


On Wed, 17 Jun 2015, Bill Broderick wrote:

> > > res=[]
> > > gnb=GNB()
> > > nf=NFoldPartitioner()
> > > permutator=AttributePermutator('targets',number_of_permutations)
> > > sl_gnb=sphere_gnbsearchlight(gnb,nf,reuse_neighbors=True,radius=3)
> > > for i in permutator.generate(ds):
> > >      res.append(sl_gnb(ds))

> > I think this should be res.append(sl_gnb(i))?

> Yes, my bad. That was a typo. It's res.append(sl_gnb(i))

sorry... coming too late to the party -- busy at OHBM... quick question -- why
don't you just use null_dist ? 

I think that you shouldn't bother much with "keep testing order intact" but
instead, if your experiment/design requires maintaining  temporal dependencies,
consider recently introduced  to AttributePermutator uattrs or chunks
strategy

        strategy : 'simple', 'uattrs', 'chunks'
          'simple' strategy is the straightforward permutation of attributes (given
          the limit).  In some sense it assumes independence of those samples.
          'uattrs' strategy looks at unique values of attr (or their unique
          combinations in case of `attr` being a list), and "permutes" those
          unique combinations values thus breaking their assignment to the samples
          but preserving any dependencies between samples within the same unique
          combination. The 'chunks' strategy swaps attribute values of entire chunks.
          Naturally, this will only work if there is the same number of samples in
          all chunks.

-- 
Yaroslav O. Halchenko, Ph.D.
http://neuro.debian.net http://www.pymvpa.org http://www.fail2ban.org
Research Scientist,            Psychological and Brain Sciences Dept.
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        



More information about the Pkg-ExpPsy-PyMVPA mailing list