Hi MNE team!

I’m doing a time-frequency analysis on the difference between two conditions with regards to activity in one EEG channel. Everything has been going well so far, but the finding (and visualizing) of significant clusters has been a bit confusing to me as I’m unsure in what format the permutation_cluster_test wants me to present the data.

Here is a quick rundown of my data:

Basically, every participant has a number of recordings in each of the two conditions. I’ve made a list of raws for each of the conditions, containing every recording that was made within a condition. I’ve then extracted the epochs for each of the recordings in each of the conditions, and averaged these out to end up with one evoked per recording per condition. I’ve then combined these evoked objects into one evoked object per condition, after which I combined these combined evokeds into a evoked for the contrast between the two. Finally, I used tfr_morlet to compute the time-frequency representation of this contrast.

This all works fine, and I can plot the contrast TFR on which clusters appear to be clearly visible. However, I am having trouble determining and visualizing the signifcant clusters. The TFR.data for each condition that I use as input for the permutation cluster test is of the shape [n_channels(1), n_frequencies(35), n_times(449)], but the documentation for the mne.stats.permutation_cluster_test function states that it wants the data for each condition to be of shape [n_observations, n_frequencies, n_times] as I understand it.

What does the observations dimension here refer to? Is it the different epochs within the condition? the evokeds of single recordings in a condition? Any help would be much appreciated!

Thanks in advance. I feel like I’m probably missing something really basic here, so apologies if the answer is very obvious!

Kind regards,

Sebastiaan