FIX: 0-index ICA channel names in get_sources

[kdoelling1919]

Tiny fix to make ICA channel names 0-indexed in the output of ica.get_sources as they are in the ICA plot functions.

@dengemann

[codecov-io]

Current coverage is 74.13% (diff: 0.00%)

Merging #3633 into master will decrease coverage by 13.27%

@@             master      #3633   diff @@
==========================================
  Files           343        343           
  Lines         60888      60761    -127   
  Methods           0          0           
  Messages          0          0           
  Branches       9307       9301      -6   
==========================================
- Hits          53220      45046   -8174   
- Misses         4912      13434   +8522   
+ Partials       2756       2281    -475   

Powered by Codecov. Last update 24ca938...f7033e9

[agramfort]

channel names for ICA are called ICA 001, ICA 002 etc.

if you want to display as number it should be displayed as ICA #0, ICA #1 etc.

[kdoelling1919]

I've updated it so that component name is 'IC #000'

It makes sense to me that both plot_components and get_sources should have the same indexing. Otherwise it can be hard to keep track if you are comparing across plots. Other channel types seem to be 1-indexed though (at least for KIT) so I see the point.

[mmagnuski]

I also like IC #000, but you should make sure that the naming is consistent across mne.
BTW @agramfort managed to reference first PR to mne python :)

[kdoelling1919]

@mmagnuski You mean that we should ensure consistency across mne with regards to IC plots, yes? Not that we should 0-index the naming for all channel types. I am a fan of 0-indexing everything but I guess other channel names are one-based indexed for a reason.

A great accidental reference from @agramfort

[agramfort]

we discussed this many times

conclusion so far is:

• name ICs as IC #0, IC #1, etc. in plots
• name channels in ch_names and info['chs'] as ICA 001 ICA 002 etc. for consistency with EEG 001, EEG 002 etc.

[kdoelling1919]

Okay fair enough. Thanks for clearing it up.

Actually, as I was trying to figure out evoked.plot functionality regarding this I couldn't actually get it to plot without error (before all this, I was just extracting the data and using plt.plot()). The error I found is described in Issue #3647

Perhaps, after that is sorted out, we can come back to this. I'm wondering what you think about creating an option for the user to change format of plot legends. Perhaps (probably) that is already implemented and I don't know about it.

[agramfort]

closing?