[ENH] Add diagonal reference line to kappa/rho plot

[notZaki]

Closes #582.

Changes proposed in this pull request:

• Add a diagonal line (y = x) in kappa/rho scatter plot to make it easier to identify which of the two is greater for each component

---

Other notes:

• Here's what the figure looks like the for the three test datasets
  
• The above scatter plots do not match the ones in the three-echo and five-echo reports from an older run
  • Assuming the older run used the same data, the differences could be due to changes in the code (Examine differences in results between original MEICA & current tedana #622) or if different input arguments were used
• Perhaps a bit more troubling: I got different results by changing the python environment. Gif below shows the scatter plot from the three-echo test using Python 3.6 (pre-installed/configured on server) and 3.8 (conda environment). 🤷
  

[jbteves]

Hi @notZaki ! Thanks so much for opening this PR. Would you mind mentioning the differences in results as an issue, since you've already started looking into it? I'll follow up with you there!

[tsalo]

This addition looks great @notZaki. Thank you.

While I think the change itself is straightforward (even knowing little about bokeh), I want to figure out why the tests did not run. It looks like your branch is up-to-date with our main branch, so the issue isn't that the CI configuration is out-of-date. I did notice that you had CircleCI set up on your fork, which I think can cause problems for CI on the main repository, but it looks like this new branch doesn't have any builds even there.

@jbteves do you have any ideas what could be happening?

[jbteves]

I'm as confused as you are @tsalo. This is basically the same behavior we see with the all contributors bot, but that shouldn't apply here because it's a code change and there's no [skip-ci] tag. Sometimes pushing an empty commit correctly triggers things (which is annoying, but we could try it, especially since we usually squash and merge anyway).

[tsalo]

Thanks @jbteves. Tagging @ME-ICA/tedana-devs then. Hopefully someone on the team will have some insight.

EDIT: Sorry, or we could ask for an empty commit.

[jbteves]

Let's try the empty commit first, just because it's easy. Would you mind either pushing an empty commit or okaying one of us to do so @notZaki ?

[notZaki]

I tried making an empty commit. Doesn't seem like it made a difference---except the docs got built this time.

If someone else wants to push an empty commit, then feel free to do so.

I'll add that I did fiddle with my circle-ci settings some time ago. Though I wouldn't expect my mangled circle-ci configuration to affect the PR.

[jbteves]

I feel like it shouldn't affect things, since you haven't modified the CircleCI config in your PR, and it should trigger the tedana CircleCI run rather than one from your repo. Bizarrely, CircleCI didn't start any jobs at all for this PR from our dashboard. Let me see if I can come up with anything that might be causing this.

[eurunuela]

It may be related to the CircleCI settings on your end @notZaki

It could be the case that it's running the tests on your fork and not on the ME-ICA repo. Could you please share a screenshot of your CircleCI settings?

[notZaki]

@eurunuela Looks like you got it right---the test were trying to run on my fork so the configuration on my end was the issue.

I deactivated/reactivated the project on my circle-ci to reset it, and looks like it's working now.

[eurunuela]

I'm glad it works now.

From my experience, ideally, your CircleCI should not have the notZaki/tedana project (I think CircleCI uses the "follow"/"unfollow" terms). This way, it would only run tests on the PRs in ME-ICA/tedana and that would avoid any issue like the one we've just had. 😉

[tsalo]

Tests are now passing (thank you everyone for figuring it out and fixing it) and the resulting reports look good to me. Approving.

[jbteves]

LGTM. Amazing, thanks @notZaki !

[jbteves]

@all-contributors please add @notZaki for code.

[allcontributors]

@jbteves

I've put up a pull request to add @notZaki! 🎉