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

[Michael Hanke]

On Mon, Mar 14, 2011 at 11:59:15AM -0400, Yaroslav Halchenko wrote:
> Hi Nynke,
> 
> quick answer (Michael might correct me): to add variable -- yes, but to
> make it present as a covariate for all the searchlights -- I do not
> think so... but should be implementable -- just would need some kind of
> proxy for QueryEngineInterface (in 0.6) which would complement
> each query for the neighboring voxels with the indexes of those
> additional features.

If you want to have that in a cheap way you can simply wrap your
searchlight'able measure with another custom measure.

Say you are doing a cross-validation in the spheres:

  cv = CrossValidation(...)

you compose a wrapper:

  m = MyMeasure(cv, covariate_ds)

that custom measure implements _call() to hstack() the incoming dataset
(which will be one particular sphere) with you covariates. Afterwards it
will run the same CV as before.

That should do it.


Michael

-- 
Michael Hanke
http://mih.voxindeserto.de