Functional connectivity methods are asymmetric

[olivercliff]

Hi all,

When using this dataset, we found that the following methods (that are named as functional on the main page) yield directed values:
coherence_phase
phase_locking_value
phase_lag_index
weighted_phase_lag_index

The dataset has brain regions on the first axis and time dimension on the second (so it should be transposed before inputting to the toolkit).

Any ideas?

[edeno]

You are correct about these measures being asymmetric. They are typically used for functional connectivity by taking the absolute value or magnitude of the signal. Since they can be used either way, I chose to leave them in their complex form to leave that choice to the user. I should be more explicit in the documentation about this.

[olivercliff]

Thanks for your response, Eric. It seems that only the phase_locking_value method is complex though (the rest are floats). Should the other ones also be asymmetric/complex?