Imaginary Coherence

Hello Sylvain & Raymundo
I wanted to ask If I want to look for imaginary coherence in a time window of around 150 ms. I have around 28 subjects with a Toal of 1700 + trials, Should I calculate coherence (N*N matrix) for each trial and then average it, or I should go for calculating coherence on a concatenated event file.

I tried calculating it for every trial and saving it as an average for each subject. The I averaged the coherence matrices using average everything option in average tab. (Referring the average tutorials). However, I am doubtful about this approach because coherence matrix is a complex valued metric.

Could you please suggest me how should I approach calculating and averaging coherence across subjects for a small time window of 150 ms. I referred previous discussions, But I'm still confused. Please suggest me any further reads related to averaging coherence matrix.

Hoping for a prompt reply.
Thanks in advance.

None of the above. Compute coherence using for all the trials in a Condition in a Subject, using the Output option: Estimate & save across combined files/epochs.

Thanks for your reply.
But my question was to calculate an average across all subjects. As I have three groups of subjects. And I want to compare coherence of these three groups.( unmatched groups)

Thanks in advance!


It all depends on your scientific question. If you want to compare between groups, you probably want to compare between samples consisting of average coherence measures for each subject in each group. Hence you don't want to average across subjects in a group but rather test for difference of coherence means between groups (for instance). You could go that by computing repeated t-tests between groups and correcting for multiple comparisons. You could do more sophisticated tests but again, it depends on your scientific question.

Thank you so much for your response.
Yes, I just wanted to know whether averaging metrics made sense. I got my answer. Thanks again.