Restore SARS-CoV-2 Forecasts

joverlee521 commented 1 year ago

Description of proposed changes

Restore SARS-CoV-2 Forecasts page + interactive app.

Testing

[x] Checks pass

joverlee521 commented 1 year ago

Text on the main forecasts page (/sars-cov-2/forecasts) should be edited within the static section of the nextstrain.org repo (/static-site/src/sections/sars-cov-2-forecasts-page.jsx).

~Text on the interactive viz page (/sars-cov-2/forecasts/interactive) should be edited within the viz app of the forecasts-ncov repo (viz/src/Panels.js).~

joverlee521 commented 1 year ago

In offline chat with @jameshadfield, we decided the easiest path forward is to only include with the static page/static images and remove the interactive viz app for now. This will give us more time to improve the viz app itself and to decide on how to move forward with replacing Gatsby to allow D3 components within nextstrain.org. It is also simpler to edit the static text just within the nextstrain.org repo and not have to work with forecasts-ncov repo as a git subtree.

This PR has been updated with cherry-picked commits from #580 that only add/update the static web page.

marlinfiggins commented 1 year ago

Lovely work! I do like the new disclaimer and the lead text. I think there's still use for a technical details page but that can wait for now.

I also agree with John and Richard that current title is a bit confusing. I would prefer switching it to dynamics as John suggests. This also highlights that the point is not restricted purely to frequency forecasts and that we're also interested in the growth advantages as well.

Also, seconding that we should have sort of line at GA=1, so that it is clear that the reference point for BA.2. growth advantage is clear.

Things I think should be changed now

Changing title to dynamics
Adding line for GA = 1

Things that should occur over next couple of weeks

Moving back to interactive viz
Toggle to show underlying frequencies (perhaps with Binomial Sampling CIs)
Toggle to logit-scale
Adding in proper forecasts and a toggle here to show them
Adding in DirMultinomial option to allow for over-dispersion in sequence count
Technical details page
(Pie in the sky) Adding in toggle for posterior predictive checks

joverlee521 commented 1 year ago

Thanks for reviewing @huddlej and @marlinfiggins!

I wish there was some text explaining why the countries that are shown were selected. Following the link to the forecasts-ncov README doesn't give me any details about this. Could that information be included in either or both of the main page and README?

I've added a Data Prep section to the forecasts-ncov README. I'm hesitant to put this in the webpage in case data prep configs change over time.

I noticed that there is a lot of white space between columns in both sections of plots. What if the overall figures (representing all three columns) were narrower or the individual panels were wider? I see why one might prefer to maintain the equal aspect ratio, so narrowing the outer figure might be better.

I believe @jameshadfield will be adjusting the white space for the larger screensizes in the viz library. (see Slack thread)

I also agree with John and Richard that current title is a bit confusing. I would prefer switching it to dynamics as John suggests.

I've renamed things from "forecasts" to "dynamics" in 9b1d7295d2bb363d471231b42d675e07046a1955.

Also, seconding that we should have sort of line at GA=1, so that it is clear that the reference point for BA.2. growth advantage is clear.

This will also have to be changed in the viz library used to generate the static images. Once updated in the viz library and we update forecasts-ncov to use the latest tarball, then the underlying static images will include the threshold line without any further changes here.

nextstrain / nextstrain.org