[pymvpa] search-light vs. ROI analysis: significance puzzle

[J.A. Etzel]

On 2/25/2013 2:17 PM, Vadim Axel wrote:
> BTW, why one should use a search-light which is substantially
> smaller than ROI? I deliberately make it of size of my ROI to make
> results more comparable between two analyses.
That isn't the suggestion I meant to make! I was just listing the
assumptions behind my advice: often, searchlights are quite a bit
smaller than ROIs. Using searchlights similar in size to your ROI is
quite sensible for comparison purposes, and avoids that issue of 
different numbers of voxels in the ROI and searchlights.

> I move my search-light each time one voxel in one of the directions,
> so each voxel participates in dozens of lights. At the end, I
> average all the predictions in each voxel.
Do you mean you're not assigning each searchlight's accuracy to its 
center voxel but rather averaging, sort of like Malin Björnsdotter's 
work? That will change things; I'd assumed you were doing a more 
"typical" analysis.

> I would appreciate if you notify me once your paper is in press.
I'll let everyone know. :)


Jo