[Neuroimaging] Performing voxel wise logistic regression

bthirion bertrand.thirion at inria.fr
Tue Nov 24 17:38:26 EST 2015


     Dear Sulantha,

     I'm not sure I understand your question: do you mean that C is a 
(categorical) scalar value (e.g. disease status of the patient or some 
task index), or C is also a volume ?
     I guess that C is a scalar, and that you're in fact running some 
kind of multivariate analysis. You may want to take a look at the 
following example: 
nilearn.github.io/auto_examples/decoding/plot_haxby_anova_svm.html#example-decoding-plot-haxby-anova-svm-py 
(just replace the SVM with a logistic regression).

     Regarding the interaction term, we don't have implementation for 
that AFAIK, but this is really straightforward to do with numpy 
pointwise mulitplication of arrays after applying the NiftiMasker.

      Best,

Bertrand

On 24/11/2015 23:27, Sulantha Sanjeewa wrote:
> Hello all,
> I would like to know whether it is possible to perform voxel wise 
> logistic regression with two volumetric regressors (with an 
> interaction term) using Nipy.
> Model:
> C ~ A + B + A*B
> Where, C is either 1 or 0,
> A and B both volumetric data.
> Thank you very much.
>
>
>
> _______________________________________________
> Neuroimaging mailing list
> Neuroimaging at python.org
> https://mail.python.org/mailman/listinfo/neuroimaging

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/neuroimaging/attachments/20151124/a29530a7/attachment.html>


More information about the Neuroimaging mailing list