[pymvpa] Sensitivity and Specificity in Searchlight analysis
debian at onerussian.com
Mon Nov 8 15:07:19 UTC 2010
quick answer would be : yes
Just construct a new measure (instead of a plain cross-validation) which
would not return only errors but also sensitivities.
but then it might get really tricky to make sense of them since they
would correspond to maps of sensitivities among the voxels in each
sphere, correspondence to original space might be lost unless you
return feature ids as well or map back into original space before
returning from the measure (but you would need to convert it to sparse
format since otherwise you would exhaust your memory if you return a
full volume for each sphere)... also combining them all afterward would
be non-trivial... you could though return just a sensitivity for the
center voxel, but that might also be difficult to interpret in
isolation, i.e. without other voxels in a sphere...
so what is the goal you are trying to achieve?
On Mon, 08 Nov 2010, Amy L. Palmer wrote:
> Hi all,
> I'm wondering if it is possible to get maps for sensitivity and
> specificity after performing a searchlight analysis in addition to the
> accuracy map.
> Pkg-ExpPsy-PyMVPA mailing list
> Pkg-ExpPsy-PyMVPA at lists.alioth.debian.org
=------------------------------ /v\ ----------------------------=
Keep in touch // \\ (yoh@|www.)onerussian.com
Yaroslav Halchenko /( )\ ICQ#: 60653192
Linux User ^^-^^ 
More information about the Pkg-ExpPsy-PyMVPA