[pymvpa] AUC brain map for searchlight

[Sean Tey]

Hi all,

I am new user of PyMVPA and I've been using it for quite a few months now.
I'm having trouble getting a searchlight AUC brain map as the output for my
analysis. When I am not running searchlight, I am able to extract AUC
values from the 'stats' conditional attribute of the CrossValidation
object. However, I'm not sure how to make searchlight use AUC values from
the 'stats' conditional attribute.

The other option I have is to set 'errorfx=mvpa2.misc.errorfx.auc_error' in
the C.V. object but that just gives me the following error:
AttributeError: 'bool' object has no attribute 'sum' in errorfx.pyc --->
https://pastebin.com/3MBMsK2M(non-searchlight-code) and
https://pastebin.com/SYXMXt43(searchlight code). I have version 2.6.5 of
pymvpa installed.

The key parts of the code snippet I am using to generate the searchlight
output is as follows:

mvpa_lsvm_clf = mvpa2.clfs.svm.LinearCSVMC(C=1)

cv_lsvm_mvpa = CrossValidation(mvpa_lsvm_clf,
                               ChainNode([NFoldPartitioner(),

Balancer(attr='targets',count=1,limit='partitions',apply_selection=True)],
                                             space='partitions'),
                             errorfx=mvpa2.misc.errorfx.auc_error,
                            enable_ca=['stats'])

sl = sphere_searchlight(cv_lsvm_mvpa, radius=3, postproc=mean_sample())
sl_result_lsvm = sl(evds_binary)

Any comments or examples of how to make an AUC brain map instead of an
accuracy brain map from searchlight is much appreciated. Thanks!

Best,
Sean Tey
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://alioth-lists.debian.net/pipermail/pkg-exppsy-pymvpa/attachments/20190416/ede08264/attachment.html>