[pymvpa] [mvpa-toolbox] Normalised Data

[Yaroslav Halchenko]

On Wed, 16 Sep 2009, Sam wrote:
> (1) Is there any reason why it would not be a good idea to carry out a
> MVPA on data that has been normalised into standard space (MNI) via
> the SPM segment/normalise functions.    Obviously I realise that the
> spatial accuracy/resolution of the analysis will be reduced due to the
> warping involved and that in principle it would be best to carry out
> the analysis in native space.  But can anyone think of a reason that
> it might inflate performance of a classifier?
quick answer: if you use searchlight later on, and you increased the
spatial sampling, therefor increased number of voxels -- you might
expect higher absolute number of false positives.

for better answer -- it depends on what you are trying to do with "MVPA"
;)

> (2) What degree of smoothing, if any, have people found to be best?

First of all it depends on what effects you are looking for.
Talking in general, there had been a relevant response from Kamitani
to Op de Beck's critique:
http://dx.doi.org/10.1016/j.neuroimage.2009.06.040

P.S. Sorry for cross-posting but I thought to mention to our list as
well ;)
-- 
                                  .-.
=------------------------------   /v\  ----------------------------=
Keep in touch                    // \\     (yoh@|www.)onerussian.com
Yaroslav Halchenko              /(   )\               ICQ#: 60653192
                   Linux User    ^^-^^    [175555]