<div dir="ltr">Hi,<div><br></div><div>Many thanks for your answers! </div><div>I would like to identify brain regions sensitive to speech intelligibility. I have already done this with GLM by comparing responses to blocks of intelligible and unintelligible sentences. However, I would also like to try if MVPA finds some other regions since I have understood that it is more sensitive. Perhaps this could be done by running searchlight analysis on the full brain and then analyzing the clusters as introduced in Etzel <a href="http://et.al">et.al</a>. (2013, i.e. the link in the previous message). </div><div> </div><div>I tried searchlight but it gives me the following warning:</div><div><br></div><div>WARNING: Obtained degenerate data with zero norm for training of <LinearCSVMC>. Scaling of C cannot be done.<br></div><div><br></div><div>I wonder if you have any advice how to solve this problem?</div><div><br></div><div>Regards,</div><div>Maria</div></div><div class="gmail_extra"><br><div class="gmail_quote">2016-01-21 17:02 GMT+02:00 Jo Etzel <span dir="ltr"><<a href="mailto:jetzel@wustl.edu" target="_blank">jetzel@wustl.edu</a>></span>:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">I quite agree with Nick's "quite tricky": about the only way in which averaging the weights over 18 the cross-validation folds will give you a correct impression of the "important" voxels is if most of the voxels in your ROI have no information at all, and the remaining are uniquely informative (each distinguishes the classes, but not correlated with each other). Needless to say, this scenario is not exactly common for fMRI datasets. (and even more complicated if multiple people are being analyzed.)<br>
<br>
Searchlights can give a decent reflection of where *local* information occurs, though there are many caveats (to cite myself, see <a href="http://www.ncbi.nlm.nih.gov/pubmed/23558106" rel="noreferrer" target="_blank">http://www.ncbi.nlm.nih.gov/pubmed/23558106</a>).<br>
<br>
I generally suggest tailoring the analysis to the hypothesis. If you're really interested in the activity in individual voxels, some sort of mass-univariate analysis is probably best. If you're interested in ROIs, ROI-based MVPA can work very well. But trying to interpret *voxels* from a *ROI-based* analysis is problematic at best.<br>
<br>
Jo<div class="HOEnZb"><div class="h5"><br>
<br>
<br>
On 1/21/2016 8:27 AM, Nick Oosterhof wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
On 21 Jan 2016, at 15:18, Maria Hakonen <<a href="mailto:maria.hakonen@gmail.com" target="_blank">maria.hakonen@gmail.com</a>><br>
wrote:<br>
<br>
I am working on my first fMRI data and would like to try MVPA<br>
analysis. I have two classes that I have classified with linear<br>
SVM. I would like to determine which voxels contribute most to the<br>
clasifier’s successful discrimination of the classes. As far as<br>
understand, the absolute value of the SVM weights directly reflect<br>
the importance of a feature (voxel) in discriminating the two<br>
classes.<br>
</blockquote>
<br>
Interpretation of SVM weights is quite tricky, see for example Haufe<br>
et al 2015 Neuroimage, doi:10.1016/j.neuroimage.2013.10.067.<br>
<br>
If you want to make inferences about the spatial location of<br>
multivariate discrimination, you may want to consider using a<br>
searchlight analysis instead.<br>
<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
I would like to average the SVM weights across all 18<br>
cross-validation folds for each voxel and wrap the resulting map<br>
into the standard space in order to display a map of the resulting<br>
overlap.<br>
</blockquote>
<br>
Even if one would be confident that SVM weights were interpretable,<br>
why take the absolute value? It would seem that this makes it much<br>
more difficult to do any stats or interpret the results. In<br>
particular, lack of signal but difference in variance of weights<br>
across regions may then yield differences in average absolute<br>
values.<br>
</blockquote>
><br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
_______________________________________________<br>
Pkg-ExpPsy-PyMVPA mailing list<br>
<a href="mailto:Pkg-ExpPsy-PyMVPA@lists.alioth.debian.org" target="_blank">Pkg-ExpPsy-PyMVPA@lists.alioth.debian.org</a><br>
<a href="http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/pkg-exppsy-pymvpa" rel="noreferrer" target="_blank">http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/pkg-exppsy-pymvpa</a><br>
<br>
<br>
</blockquote></div></div><span class="HOEnZb"><font color="#888888">
-- <br>
Joset A. Etzel, Ph.D.<br>
Research Analyst<br>
Cognitive Control & Psychopathology Lab<br>
Washington University in St. Louis<br>
<a href="http://mvpa.blogspot.com/" rel="noreferrer" target="_blank">http://mvpa.blogspot.com/</a></font></span><div class="HOEnZb"><div class="h5"><br>
<br>
_______________________________________________<br>
Pkg-ExpPsy-PyMVPA mailing list<br>
<a href="mailto:Pkg-ExpPsy-PyMVPA@lists.alioth.debian.org" target="_blank">Pkg-ExpPsy-PyMVPA@lists.alioth.debian.org</a><br>
<a href="http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/pkg-exppsy-pymvpa" rel="noreferrer" target="_blank">http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/pkg-exppsy-pymvpa</a></div></div></blockquote></div><br></div>