[Mne_analysis] permutation_cluster_test precision

Pieter Huycke Pieter.Huycke at ugent.be
Mon Mar 23 11:39:52 EDT 2020
Search archives:

        External Email - Use Caution        

Dear MNE community,

I have been using the function mne.stats.permutation_cluster_test<https://mne.tools/stable/generated/mne.stats.permutation_cluster_test.html> as a statistical test to compare clusters in time-frequency power estimates between conditions.
Then, I plot the clusters that survived the cluster correction on top of the original data, in line with the docs.

The results make sense, but I have the idea that the cluster is quite ‘spread’.
Specifically, the surviving cluster spans multiple frequency bands, and extends pretty far in time.
Is there a way to make the resulting cluster more restricted, both in the time- and frequency domain?
I can, for example, now mask the resulting image and only keep the resulting, more significant ‘blobs’, but this is of course not as clean as an algorithm that outputs these blobs without extra steps.

Perhaps relevant: the parameters I used so far (with d1 being the data in condition 1)
mne.stats.permutation_cluster_test([d1, d2],
                                                         n_permutations = 1000,
                                                         step_down_p    = 0.05,
                                                         threshold           = None,
                                                         out_type            = "indices",
                                                         verbose              = True)

Thank you,

Kind regards,

Pieter
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20200323/4f86f9a5/attachment.html 


More information about the Mne_analysis mailing list