[pymvpa] Partial least square(PLS) analysis

[Zhen Zonglei]

Thanks for per and yark.

We know that Rotman toolbox is a good candidate to do PLS analysis in
neuroimaging, but we always hope we can do it in pymvpa framework because it
is simple and easy to use.

In our situations, generally, there are two groups of variables, one group
is multi-voxel/vertex/region pattern, and the other group is a set of scores
from a group of behavior tests. Our target was to test whether the two group
variables are relevant (correlated) with PLS, i.e., whether the neural
patterns can predict these behaviors. If they show high relevant, we also
want o know what are the contribution of each voxel and behavior score to
the correlation. Furthermore, we will do cross-validation for the PLS model.

The multi-voxel/region pattern may derive from MRI, fMRI, rfMRI or DTI. For
example, we can test whether the thickness pattern of brain can predict
subjects’ attention abilities which are test by a group attention paradigm
such as flanker, alert, top-down and bottom-up test.

For PLS applications in neuroimaging, there is a good tutorial to start. The
link as bellow:

A Krishnan, et al, 2010. Partial Least Squares (PLS) methods for
neuroimaging: A tutorial and review, NeuroImage.

http://linkinghub.elsevier.com/retrieve/pii/S1053811910010074





Zonglei


2011/5/12 Yaroslav Halchenko <debian at onerussian.com>

> Hi Per, Hi Zhen,
>
> actually I think things are not that hopeless ;-)
>
> PLS has been implemented in scikits.learn, so, most probably, there is
> no need to reimplement it, unless you see some obvious drawbacks in
> scikits.learn implementation(s)?  on PyMVPA side we have started
> implementing a basic adapter to make scikits.learn machinery
> easily pluggable within PyMVPA.  So, theoretically, you could already
> started playing with something like
>
> from scikits.learn.pls import PLSRegression as sklPLSRegression
> pls = SKLLearnerAdapter(sklPLSRegression())
>
> so pls should look like a regular PyMVPA learner.
>
> But could you please very briefly describe what kind of a task you are
> looking for to tackle with PLS, e.g. what would be your "targets" and where
> within processing pipeline PLS would need to be used?
>
> On Wed, 11 May 2011, Per B. Sederberg wrote:
> > Hi Zonglei:
>
> > I started to implement PLS in PyMVPA now almost 3 years ago, but never
> > finished.  I still think it would be a good addition, but I don't have
> > time to finish it up.  I'd love to replicate a good bit of the PLS
> > functionality from the Rotman toolbox:
>
> > http://www.rotman-baycrest.on.ca/index.php?section=84
>
> > So, to my knowledge, there is no way to do it now, but I'd be happy to
> > chat with someone who would like to take on finishing up this
> > addition.
>
> > Best,
> > Per
>
>
> > On Wed, May 11, 2011 at 9:48 PM, Zhen Zonglei <zonglei.fsl at gmail.com>
> wrote:
> > > Dear all:
> > >    Is there a easy way to do PLS analysis in Pymvpa?  I exepect to
> > > investigate the correlation between neural patterns and a group of
> behavior
> > > index, so I restort to PLS.
> > > If you have any info on how to do this in pymvpa, please let me know.
>
> --
> =------------------------------------------------------------------=
> Keep in touch                                     www.onerussian.com
> Yaroslav Halchenko                 www.ohloh.net/accounts/yarikoptic
>
> _______________________________________________
> Pkg-ExpPsy-PyMVPA mailing list
> Pkg-ExpPsy-PyMVPA at lists.alioth.debian.org
> http://lists.alioth.debian.org/mailman/listinfo/pkg-exppsy-pymvpa
>



-- 
Zonglei Zhen, PhD
BNU Imaging Center for Brain Research, State Key Laboratory of Cognitive
Neuroscience and Learning, P.R. China
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.alioth.debian.org/pipermail/pkg-exppsy-pymvpa/attachments/20110512/0fa0cdaf/attachment.htm>