Update: The COVID tracking project paused their data API on March 7, 2021.
TLDR: Link to the dashboard is here.
I made a dashboard to visualize COVID-19-related exponential growth factors and multi-state level trends in the US. It’s the beginning of May, and most states are thinking of reopening this month, so I figure now may be a good time for people to discover another set of tools to keep an eye on our collective suffering.
For a quick look on what it can do, here’s a link to the dashboard for the 10 states with the highest test coverage as of now, with the chart plotting the growth factor for the combined number of deaths.
In the US, the currently available data are essentially of 3 types: the number of tests conducted, the number of positive cases, and the number of deaths. This dashboard presents both per-capita figures and the change of them over time.
Hospitalization-related figures and especially hospital capacity figures would be very helpful, but generally they are of dubious quality and missing in many states. I’ve decided at least now to ignore them in my analysis. My belief is that with bad legs it’s often better to just trudge forward painfully than to risk relying on unreliable crutches.
I won’t go into any interpretation of specific datasets here, though I do have a few preliminary thoughts on the available types of data. Out of the 3 types (tests, positive cases, and deaths), the death-related figures are the most concrete despite being lagging indicators; this makes them the most reliable candidate to be used as the prior for deductive reasoning. The growth factor of the positive cases is the most straightforward figure to use for outbreak detection, but its effectiveness is plagued by the existence of asymptomatic cases and the overall lack of testing done in the country. Even though the absolute number is low, the growth factor of the number of tests conducted is a great proxy for the state and federal governments’ effectiveness or the lack thereof.
The usage of other forms of the 3 types of data seems to require non-obvious justification and research, but I would love to find out more about them.
And with all that out of the way, here are some interesting groups of states to check out and potentially bookmark:
I’d love to see more interesting ways of grouping the states, so feel free to play around with the dashboard and get in touch on Twitter. The source code is hosted on GitHub, and all types of contributions are welcome.