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

[Francisco Pereira]

On Mon, Mar 14, 2011 at 1:49 PM, Yaroslav Halchenko
<debian at onerussian.com> wrote:
>
> I guess you mean that if effects of the additional variables are not
> additive (and most often they would not be), then they indeed would be
> of little help for the linear classifier.

Much better put than I did, thank you!

> On the other hand, if additional variables themselves covary with the
> targets of interest, they would impact ALL the voxels, thus adding a
> global bias to the results across the voxels.

Right, I can see that. The best scenario I had in mind would be
something like CART, where a categorical variable divides the examples
and then there is a separate classifier/regression for each partition.

Francisco