[pymvpa] Combine the fMRI data with other data in the pattern analysis?

Yaroslav Halchenko debian at onerussian.com
Wed Mar 16 13:52:42 UTC 2011


On Wed, 16 Mar 2011, Nynke van der Laan wrote:
> the right way. See below for the total script (could be useful for
> other users with version 0.4 that want to do the same). Would you

cool -- seems to come out quite neat and concise, isn't it?

> please check it if this is indeed what it is doing?

;) you tell us -- is there an effect on the results in comparison to
plain searchlight on cv?  to me it looks ok... see more of inline
comments below though

> AFTER THAT, the covariate is added as an additional feature? So for

yes! and you could see that easily if you just 

print ds.nfeatures

within your __call__ to see that you are getting 'spheres'

> to each sphere). Then I wonder: doesnt' the fact that the covariate
> has very different values (i.e., in the range of 1-9 instead of the
> much smaller voxel values) complicate the analysis?

it will.  Just zscore (or normalize them any other way) prior  giving to
MyMeasure .

> class MyMeasure:
> 	>...<
> 	def __call__(self, ds):

although it should be perfectly ok for your case, you might have wanted
to make

class MyMeasure(DatasetMeasure):
...
  def _call(self, ds):
    ....

in case if you wanted to assign any transformer to be
applied... but in your case it should be just fine as is.

-- 
=------------------------------------------------------------------=
Keep in touch                                     www.onerussian.com
Yaroslav Halchenko                 www.ohloh.net/accounts/yarikoptic



More information about the Pkg-ExpPsy-PyMVPA mailing list