Hi Nynke, Without wishing to make a "commercial" pitch, Searchmight has that as an (undocumented) feature because of local requests. If it's something you desperately need I can help you use that. It shouldn't be hard to implement, though there are issues with making sure that the range of values the extra variable takes are comparable to those from the voxels (judging by our experience). A different consideration is whether the classifier can use the value of the extra variable to change the way it uses the others; at first thought I'm not sure a linear classifier will do there (but could well be wrong).
cheers, Francisco On Mon, Mar 14, 2011 at 9:49 AM, Nynke van der Laan <[email protected]> wrote: > Hello, > I have a question concerning searchlight analyses: Is there a build in > possibility in pyMVPA to add an extra variable to the pattern > analysis? E.g., if you have for each sample, next to the neuroimaging > data also skin conductance measures (1 per sample/scan). Would it be > possible to add this variable as a covariate to be able to see whether > this variable can improve the prediction accuracy? So when you run an > analysis with only the functional scans, and one with both the skin > conductance and the functional scans, you can see whether that adds > any information. > Thanks in advance for answering my question! > Best regards, > Nynke > > _______________________________________________ > Pkg-ExpPsy-PyMVPA mailing list > [email protected] > http://lists.alioth.debian.org/mailman/listinfo/pkg-exppsy-pymvpa > _______________________________________________ Pkg-ExpPsy-PyMVPA mailing list [email protected] http://lists.alioth.debian.org/mailman/listinfo/pkg-exppsy-pymvpa
