Hi, On Mon, Sep 28, 2015 at 4:42 PM, Richard Dinga <ding...@gmail.com> wrote:
> > I'm performing searchlight support vector regression, not classification, > > so my error goes from 0 to 2 instead of 0 to 1. Can I simply take `2 - > > error` (so higher is better) or does the algorithm require the values to > > lie between 0 and 1? > > You can do that, as long as higher is better. Algorithm will just sort them > and find nth biggest value. > > > The docs say the accuracy maps must be the result of > > classification, but is there a specific reason regression won't work? > > I have no idea :) Original paper doesn't mention regression. Will p-values > and FDR still make sense? > > Right now I cannot think of a reason we it wouldn't work -- as long as "bigger is better". The algorithm converts pretty much anything into probabilities/frequencies. 1. Value corresponding to a particular probability under H0 -> threshold. 2. frequency with which a particular blob size has been observered in a group average map under H0. Should work. Michael -- Michael Hanke http://mih.voxindeserto.de
_______________________________________________ Pkg-ExpPsy-PyMVPA mailing list Pkg-ExpPsy-PyMVPA@lists.alioth.debian.org http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/pkg-exppsy-pymvpa