I am carrying out spectral connectivity analysis of resting state data using mne-connectivity 0.5.0 and MNE 1.6.1.
So far I am using the exact method found in the mne-connectivity documentation here: Link
I have two groups of participants that I would like to compare, which is what I would like to discuss.
Generating Connectivity Values
Initially, two approaches came to mind:
Concatenate all epochs of participants from each group using make_fixed_length_epochs and generate connectivity values for each group from these.
Generate connectivity values for each participant from fixed length epochs, then mean values of all participants in each group.
Are there any issues with generating connectivity from concatenated epochs of different participants?
This would be carried out for each frequency band.
Comparing Connectivity
Once we have connectivity arrays for each group, the question is of how to go about comparing them.
My initial thought would be to a connectivity circle plot of differences between the two groups.
Does a pairwise T-test of each brain regions’ connectivity make sense in this context?
Overall, I’m interested in exploring how to statistically differentiate these two groups using the 68x68 connectivity matrix.
Generating Connectivity Values
Initially, two approaches came to mind:
Concatenate all epochs of participants from each group using make_fixed_length_epochs and generate connectivity values for each group from these.
Generate connectivity values for each participant from fixed length epochs, then mean values of all participants in each group.
Are there any issues with generating connectivity from concatenated epochs of different participants?
Approach 2 would be the traditional way.
Keep in mind also that for any statistical comparisons you’ll probably be using the results from each participant as the different observations for your two groups, so combining everything together in approach 1 would not allow for this.
Comparing Connectivity
Once we have connectivity arrays for each group, the question is of how to go about comparing them.
My initial thought would be to a connectivity circle plot of differences between the two groups.
Does a pairwise T-test of each brain regions’ connectivity make sense in this context?
This also means that averaging the results over all participant is fine for visualising in a connectivity circle, but you would want the un-averaged results for each participant in the statistical comparison.
With the connectivity results for each participant, a pairwise T-test for each connection would be fine, but then you should also correct this for multiple comparisons (e.g., Bonferroni, FDR).
Thank you for your excellent reply. Agreed on both points, it makes sense to carry out participant-wise connectivity and I’ll apply correction for multiple t-test comparisons.