[mne-python] spatio temporal clustering stats

✉️ Mailing List Archive (read-only)
1

Hi everyone,

I'm currently trying to adapt the python spatio-temporal clustering stats example to my own data.
As a sanity check I started out with 168 right versus 168 left button presses, irrespective of the condition.
Those yield the expected left-hemispheric blob when averaged over time slices and subjects (dSPM with regular default parameters), that's good.
However, I'm now having a hard time to understand why the clustering test's p_values uniformly equal to 1.0. To me this does not seem like a 'natural' result. I hope I missed something obvious and trivial, but currently I don't see what it might be.
Any pointers would be highly welcome.

Cheers,
Denis

Here's an excerpt of the code I currently use:
https://gist.github.com/dbacc6e4941ca1082540

Also please find the average dSPM visualizations attached.

2

Apparently the figures got lost somewhere,

here you can get them:

https://www.dropbox.com/s/zyr3kuljn26firp/motor_sanity_check_lh.png
https://www.dropbox.com/s/8ukmabsstu457zw/motor_sanity_check_rh.png

Denis

3

Hi Denis,

what do you mean? cluster_p_values are all 1?

can you share the files somewhere?

Alex

4

Hi Denis,

I have my equivalent Matlab code for spatio-temporal clustering, let me
know if this some thing you like to try.

Sheraz

5

Hi Denis,

what do you mean? cluster_p_values are all 1?

Sorry, forgot to pick up this one. Yes, indeed, all p-values are 1 and hence no valid clusters. This is confusing, given the pronounced motor response and the expected high sensitivity of spatio-temporal clustering...

can you share the files somewhere?

Alex

The information in this e-mail is intended only for the person to whom it is
addressed. If you believe this e-mail was sent to you in error and the e-mail
contains patient information, please contact the Partners Compliance HelpLine at
http://www.partners.org/complianceline . If the e-mail was sent to you in error
but does not contain patient information, please contact the sender and properly
dispose of the e-mail.

6

Dear list, just to give you a short update.

It turned out there was a silent bug in 'mne.stats.spatio_temporal_cluster_1samp_test' regarding the parameter-handling for parallel computation.
background: negative values indicating the number processors *not* to be used interfered with the internal job dispatch. As a consequence the test just was not performed leaving the array of p-values unmodified.

This issue is fixed in the current development version of mne-python.

Once more thanks for your immediate responses!

Best,
Denis