spectral_connectivity_time(faverage=True) averages over roughly twice the requested band when a band does not end at the top of the frequency grid

Support & Discussions
1

In mne_connectivity/spectral/time.py, _foi_average contains

f_e += 1 if f_s == f_e else f_e

Since Python parses this as f_e += (1 if f_s == f_e else f_e), a band whose start and end bins differ gets its end index doubled instead of incremented. With freqs = np.linspace(1, 45, 133) and fmin=(1, 8, 13), fmax=(4, 13, 30), the delta average covers 1 to 7 Hz, alpha 8 to 24.7 Hz while you expect 8 to 13 Hz.

Introduced in commit 4476efae (PR #104), released in 0.5 (2023-01-14); still present on main at the same line. spectral_connectivity_epochs has its own averaging and is not affected. Is this a real bug or am I missing something?

Thank you!

Marieke

2

Hi Marieke,

Welcome to the forums, and thanks for raising this!

That’s not the intended behaviour, and should be fixed. I’ve opened an issue to track this.

A PR would be welcome, if you are comfortable submitting one. Otherwise, I could pick this up.

Cheers,
Thomas

3

Hi Thomas,

Thanks a lot for your quick reply and for confirming. I opened a PR: [MRG] Fix faverage end index in spectral_connectivity_time by mariekedevylder · Pull Request #479 · mne-tools/mne-connectivity · GitHub . It applies the change you proposed, adds a test comparing faverage=True against the mean of the per-frequency output, and includes the changelog entry. It is my first contribution here, so please let me know if I missed anything.

Thank you!

Marieke