[pymvpa] classification based on individual parameter estimates from FSL

Nick Oosterhof nikolaas.oosterhof at unitn.it
Fri Aug 1 12:31:06 UTC 2014

On Aug 1, 2014, at 11:29 AM, David Soto <d.soto.b at gmail.com> wrote:

> Thanks for the response, I have not managed to extract the whole-brain classification map...following the 1st example code below, the output from the crossvalidation is
> Dataset(array([[ 0.35526316],
>        [ 0.35855263]]), sa=SampleAttributesCollection(items=[ArrayCollectable(name='cvfolds', doc=None, value=array([0, 1]), length=2)]), fa=FeatureAttributesCollection(items=[]), a=DatasetAttributesCollection(items=[]))

To be clear, you seem to have two accuracies (or whatever value you computed here), corresponding to two folds. Both folds have used data from all features.

> How can i extract the whole brain classification map?

There is no whole-brain classification map. Your two accuracies indicate how well the conditions can be distinguished using all features. These values do not tell you /where/ in the brain conditions can be distinguished. In other words, you've lost all spatial selectivity (see my earlier message).

To get a brain map (one value for each voxel), you have at least two options:

1) make a sensitivity map: http://www.pymvpa.org/examples/sensanas.html
2) run a searchlight.

Personally I prefer option (2) because option (1) may obscure some informative regions if there are other, 'more' informative, regions as well.

More information about the Pkg-ExpPsy-PyMVPA mailing list