bug(evalcast): remove title from labs in plots

[dshemetov]

closes #640

• removes title argument from the labs
• also updates the DESCRIPTION file with remotes that work with devtools, remotes, and pak

[brookslogan]

Thanks; do you know what the plot labels look like now in the vignette?

[dshemetov]

They looked unchanged to me 🤷, but I didn't look suuuper closely

[brookslogan]

Looks good! This should fix vignette rendering and some user bugs. We might want to revisit in another Issue if the guide labels are unclear or eating up a bunch of space now [we want a title].

[brookslogan]

If you look at the wedge plots here, for example, the text "Proportion below" and "Proportion above" labeling the colors should have changed. It might be a better hack to just say "Proportion below" and "Proportion above" instead of the corresponding sprintf failures; I'm not sure what it's going to look like by default.

[brookslogan]

Sorry, never mind, my bad. There shouldn't be a change; sprintf with NULL returns character(0) [and I assume title = character(0) does nothing as there is no default title]

[dshemetov]

This is what the first plot looks like 🤷

[brookslogan]

Yep, it makes sense now; I thought NULL acted like "" in sprintf but it actually acts like character(0) [this is more natural as length(NULL) is 0, but I thought it sprintf was less natural], so we're going from buggily saying "no title" ----> not saying anything about the title & getting no title.

[dshemetov]

Once CI finishes, I'll merge. I don't think we should continue adding features to this repo, so I wouldn't open a title issue.

[zterner-mitre]

Thanks for all your work on this!! I really appreciate it. I'd be happy to confirm that things work on my end once the changes are finished, just let me know how to proceed at that time if you can.

[dshemetov]

Thanks @zterner-mitre! Can you try running through the tutorial again after installing with

pak::pkg_install("cmu-delphi/covidcast/R-packages/evalcast@ds/evalcast-plot")

[zterner-mitre]

The first plot_calibration is giving this warning and this plot, which I think is probably not what you are aiming for?

The second plot_calibration seems good:

Third also seems good

[zterner-mitre]

plot_coverage and plot_width are producing too many plots given the setup of the plots (all in one row, for instance), and also may be missing some data. I can share those with you here, or open another issue, whichever you prefer

[dshemetov]

What does your predictions_cards look like? My first plot does not match yours.

[dshemetov]

I think the issue might be that your plots are using the full forecast_dates instead of the reduced forecast_dates_dec

[zterner-mitre]

Yep! That's it. I had manually defined forecast_dates_dec yesterday when I was having issues with get_covidhub_forecast_dates to be all dates within December. But using the sequence of 4 dates instead of 31 fixed everything. Thanks!

[dshemetov]

Excellent!