Mne-python: What does set_eeg_reference() do?

[nbara]

Hi all,

I'm a bit puzzled as to what set_eeg_reference() actually does.

Here is a gist where I do set_eeg_reference(ref_channels=[]) (as my data was already referenced).

https://gist.github.com/nbara/972eefa4a62fd6ef9df308b76ffa2559

Contrary to my expectations, this seems to rereference to the mean, and also appears to include the STIM channel in the mix (visible peaks every .5 seconds)??

[larsoner]

Based on the docstring it seems like doing [] should indeed not affect the data, but clearly it did. So I think this is a bug.

Can you take a minute to use np.random.RandomState(0).randn(...) as the data to turn this into a minimal example?

Are you up for digging into the code to try to fix it?

[nbara]

Hi Eric,

Here's a minimal example using random data:
https://gist.github.com/nbara/8e6e2a3dbb18675b7b57fca3fbfbc111

Does that work?

I'm afraid I have too much on my plate atm so I won't be able to fix this myself...

[wmvanvliet]

What the.... the referencing function should not be doing that. ok, I'm looking into it.

[wmvanvliet]

Turns out, the unit test for this case was broken. This allowed me to at some point introduce a breaking change to the set_eeg_reference() function. My apologies.