[Mne_analysis] edge effects in computation of induced power?

[Alexandre Gramfort]

hi Andy,

> I've been working recently with the Python toolbox trying to examine
> higher-frequency activity in an auditory paradigm.  The PLV seems OK, but
> there are strange things happening at the edges of the induced power matrix
> which I think are obscuring whatever power is induced by the stimulus.  When
> I process the sample dataset with the exact same script, I don't see such
> edge effects.  Does anyone have any ideas about why this  is happening?

what you see is normal as no correction is done about edge artifact. The
way I do it is by clipping the TFR maps at the beginning and at the end.
Also baseline should not include the coefficients with artifact.

> A related general question: Is the evoked field subtracted from the
> individual epochs prior to computation?

no. Separating evoked from induced by subtracting the mean is usually
not enough.
I agree that the term "induced" can therefore be misleading as
currently described.

Feel free so send us some suggestions and code to improve this part of
the mne-python
code.

Best,
Alex