[pymvpa] searchlight demo clarification

[Michael Hanke]

On Fri, Dec 05, 2008 at 11:24:40AM -0500, Mark Eckert wrote:
> It looks like we have a software dependency issue that will take a little
> time to sort out.  In the mean time, maybe we can discuss the figure you
> sent.  Am I correct in interpretting that the color scale in the image
> reflects the percent of correct classification for each voxel?
It shows indeed the fraction of correct classifications -- but for the
sphere (of a given radius) centered around the respective voxel.

> Other than
> presenting a basic example of how the data can be represented, is this
> example designed to emphasize a particular point (e.g., significance of
> searchlight radius).

No, this is not intended. It simply shows that it makes a difference
which radius to choose. This can also be considered as one of the
weaknesses of the searchlight approach -- what is the 'best' radius? But
for sure this example is not suitable to answer this question.

Michael

-- 
GPG key:  1024D/3144BE0F Michael Hanke
http://apsy.gse.uni-magdeburg.de/hanke
ICQ: 48230050