[pymvpa] Suspicious results

[Nynke van der Laan]

Hello all,
I've used the toolbox to analyze my data and I get kind of strange results.
What I did is the following: I did a searchlight analysis (radius 10
mm), with the classifier RbfNuSVMC on the whole brain (with a whole
brain mask). I used a NFoldCrossvalidation (no detrending or
z-scoring).
I use two stimuluscategories. The task I used consisted of 38 chunks
(38 trials) with in each chunk two stimuluspresentations (one of each
category). I have used blockaveraging to reduce features.
My results are in the form of 3-d accuracymaps (i.e., for each voxel I
have the mean True pos + true neg / total over the crossvalidation).
Because I have two stimuluscategories the chance level accuracy would
thus be 0.5
What I get as results is that almost all voxels have an accuracy above
0.5 and the highest accuracies I got were around 0.7 (i.e., 70%
correctly classified) So this would mean that there is predictive
information in all regions of the brain.. Also, when I put all the
single subject results into a group analysis I find that all brain
regions predict above 50% accuracy.
The highest peaks are located at the borders of the brain.
My question is: Are these results possibly correct? Is it possible
that all brain areas are predictive? Or is there something strange
going on? (overfitting maybe?)
Best regards, Nynke van der Laan