search

neherlab / covid19_scenarios

Models of COVID-19 outbreak trajectories and hospital demand
https://covid19-scenarios.org
MIT License
1.36k stars 354 forks source link

Feat/weekly deaths #731

Closed rneher closed 4 years ago

rneher commented 4 years ago

Description

As we progress further into the pandemic, total deaths are not very useful to assess current trends. This PR adds weekly deaths to the graph and also changes the interval for cases to weekly. This has the benefit of smoothing out weekday variation. Need to adjust colors.

Impacted Areas in the application

plot and plotdata preparation.

Testing

do we like it?

vercel[bot] commented 4 years ago

This pull request is being automatically deployed with Vercel (learn more). To see the status of your deployment, click below or on the icon next to each commit.

🔍 Inspect: https://vercel.com/covid19-scenarios/covid19-scenarios/deyacywc8 ✅ Preview: https://covid19-scenarios-git-feat-weeklydeaths.covid19-scenarios.now.sh

ivan-aksamentov commented 4 years ago

@rneher Yes, this indeed gives some better feeling for the trend.

Model's weeklies don't display uncertainty. Bug or feature? Also, the tooltip text casing in inconsistent.

ivan-aksamentov commented 4 years ago

Patched the text casing.

rneher commented 4 years ago

the missing uncertainty is a bug... but getting this uncertainty correctly is actually a little tricky algorithmically.

ivan-aksamentov commented 4 years ago

Should we also include the number to the results and emit it into the exports? It looks somewhat easy to derive right now, so maybe not. But if the algorithm for uncertainty will be tricky, we may want to include it, for consistency.

rneher commented 4 years ago

Should we also include the number to the results and emit it into the exports? It looks somewhat easy to derive right now, so maybe not. But if the algorithm for uncertainty will be tricky, we may want to include it, for consistency.

might be a good idea. The issue with the upper and lower is that weekly increases of the lower cumulative death curve might be higher than the weekly increase of the upper, thereby screwing up what is min and max. so maybe we just have to bite the bullet and compute this along with the result.

codeclimate[bot] commented 4 years ago

Code Climate has analyzed commit c7a3aa1e and detected 0 issues on this pull request.

View more on Code Climate.

rneher commented 4 years ago

I fixed some of these issues and made a separate list of plots that should be enabled by default.