Add ERDS topomap plotting

[hofaflo]

To-Do:

• use dialogs.utils.select_all once Add support for plotting topomaps of evoked potentials #277 is merged

There's a lot of code duplication in dialogs/erds.py, but since we want to add significance masks to ERDS maps soon, I'd postpone refactoring to when that is done.

[cbrnr]

Do you want me to merge this one or is there something you wanted to refactor?

[hofaflo]

This one's ready, I'd prefer to refactor once you are happy with #279 interface- and functionality-wise :)

[cbrnr]

I just tested this PR and I have three comments:

1. The menu entry should be disabled if there are no locations (montage) associated with the data.
2. What is calculated here? The mean of the frequency and time range? Why is step size important then (for the user)?
3. How are the plots scaled? The colorbar is labeled with "AU", but since we're computing ERDS it should be labeled as such. Also, the lower boundary should be -100% (or -1, I don't even know if we're showing values in percent or not). And the upper boundary is usually between 2.5 and 3.5, maybe this should be settable (with an option to determine the maximum from the data)?

Do you have any good example data where we can test this? Maybe the data from the ERDS maps tutorial? But here the channel labels end with dots and there's no easy way to remove them in MNELAB...

[hofaflo]

I'm lacking ERDS knowledge, so I think it would be best if we discuss this on tuesday in "person" 😅

I used the data from your BCI-event ERDS Tutorial for trying things out.

[hofaflo]

The cmap used in viz.plot_erds_topomaps is now the same as in viz.plot_erds. #299 will change viz.plot_erds to use cnorm. This is not applicable here as mne.viz.plot_tfr_topomap does not have a cnorm parameter 🤔

[cbrnr]

Then I guess we should include center_cmap in MNELAB for now. The behavior of TwoSlopeNorm has changed in recent Matplotlib versions anyway, the colorbar has a weird scaling which I don't like (see my question here).

[cbrnr]

The dialog is also missing edits to select the time range that should be averaged.

[hofaflo]

The dialog is also missing edits to select the time range that should be averaged.

Strange, it's there for me 🤔

[cbrnr]

Maybe I forgot to pull. I though VS Code does this automatically.

[cbrnr]

No wait, it's not there! The time range is for computing the complete map. We need another time selection for the average.

[hofaflo]

The map is always computed with the entire epoch duration (there's no time parameter here). The time range that's entered in the dialog is then applied here.

[cbrnr]

I see, OK, that makes sense.

[cbrnr]

Do you want to include center_cmap in this PR?

[hofaflo]

Do you want to include center_cmap in this PR?

I'm not sure what you mean 🤔

[cbrnr]

Include it in MNELAB because it will be gone from MNE.

[hofaflo]

Aaaah 🙈 done!

[cbrnr]

Thanks @hofaflo!