[pymvpa] Sensitivity analysis with GNB

[Thomas Nickson]

Yeah, I see. Thanks for clarifying. So if I want to use the GNB as a form
of feature selection I can use the means to see which are the most
important to the classifier? I can just get absolute values and then look
at the top x% or something or do I need to take in to account the variance
in some way?

On Mon, Dec 15, 2014 at 4:45 PM, Nick Oosterhof <
n.n.oosterhof at googlemail.com> wrote:
>
> On 15 Dec 2014, at 17:20, Thomas Nickson <thomas.nickson at gmail.com> wrote:
>
> > Oh wait. I think you just mean that I can calculate the f-test of the
> dataset and this will give me the sensitivity measure, regardless of the
> result of the GNB classifier. Is this correct?
>
> Indeed, that’s the idea.
> For clarity, this won’t work with just any classifier, but it will work
> for the GNB because of its ‘naive-ness’.
> _______________________________________________
> Pkg-ExpPsy-PyMVPA mailing list
> Pkg-ExpPsy-PyMVPA at lists.alioth.debian.org
> http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/pkg-exppsy-pymvpa
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.alioth.debian.org/pipermail/pkg-exppsy-pymvpa/attachments/20141215/698be589/attachment.html>