Coherence values

I have 96 channel EEG data.

When I compute channel-based magnitude-squared coherence between C3-C4, I get coherence values around 0.3.

When I compute source-based magnitude-squared coherence between precentral L and precentral R, I get coherence values around 0.04.

Why is the source-based coherence so much lower than channel-based? Is it because the source-based coherence is an average of coherence values between all individual vertices, as opposed to combined signals for channels?

When I compute source-based imaginary coherence between precentral L and precentral R, I get coherence values around 0.003.

Is imaginary coherence so much lower than magnitude-squared coherence, because it eliminates the effects of the volume conduction?

Why is the source-based coherence so much lower than channel-based? Is it because the source-based coherence is an average of coherence values between all individual vertices, as opposed to combined signals for channels?

Yes, probably.

The two electrodes C3-C4 coherence are close, the potentials recorded are similar (EEG scalp topographies are very smooth) and therefore the coherence value is high.

In source space: you average a lot of heterogeneous values together, therefore your signal of interest gets lost.
Note that the "precentral" ROIs in the FreeSurfer atlas are probably too large to be used as functional regions. Try to target better the part of the motor cortex you are interested in.

When I compute source-based imaginary coherence between precentral L and precentral R, I get coherence values around 0.003.

Is imaginary coherence so much lower than magnitude-squared coherence, because it eliminates the effects of the volume conduction?
[/quote]

Is imaginary coherence so much lower than magnitude-squared coherence, because it eliminates the effects of the volume conduction?

Yes, Imaginary Coh takes lower values than squared Coh, specially when electrodes are close each other, actually, nolte (2004) does not recomend to use imaginary Coh itself as a connecivity measure but use it to build another measures, due to error II type . So you can use another metrics like: Phase Slope Index or Weighted PhaseLagIndex (both are based on imaginary Coh).

Does Brainstorm allow users to obtain Phase Slope Index or Weighted PhaseLagIndex?
I see "Phase locking value" in the Process selection menu. But I don't see Phase Slope Index or Weighted PhaseLagIndex?

No, these measures are not available yet.

@Sylvain @hossein27en: is there a plan for implementing this?

Not our priorities at this time.