Adam Li

FYI to fix the style easily and the sorting of imports: `black ./mne_icalabel/*` and `isort ./mne_icalabel` commands, or type `make black` and `make isort`. The other checks like flake8 (style),...

Hi @witherscp thanks for opening up an issue! Checking matrix rank would not be sufficient as there is a linear algebra result "almost every matrix is diagonalizable" (i.e. full rank)...

> If I check condition number using np.linalg.cond(), how can I know whether it would be considered too large? In my sample data, when there is no common average reference...

Sorry for the delay @alexrockhill . I was out of country on vacation till last night. So I think overall this is a useful feature since the time-shuffled idea works...

> Oh super interesting, so maybe we should stick with this. If we do that, would you mind writing some of the interpretation? Sounds good! Can do. Ping me when...

@alexrockhill how's this going? Are my additions sufficient for what you were thinking?

True, but that's a limitation in general for correlation based connectivity analysis. I think as long as the assumption of stationarity is said, then the time shuffling is theoretically sound...

Hey @alexrockhill sorry for the late reply. Statistical testing with especially non-symmetric connectivity matrices is a pretty challenging problem, so I'm not 100% sure how I would go about this...

Yes that is a good idea. A PR is welcome! :) Lmk if you need guidance

@Div12345 apologies for the delay. It is perhaps worth tabling this for now and actually it seems that there is some discrepancy as to if the implementation I ported over...