Read on Twitter
Total Confirmed Coronavirus Cases (% of State Population) for all 50 states
(Log-Y)
[data via @JohnsHopkins and @WHO March 11 2020]
#CoVID19
1/?
![Total Confirmed Coronavirus Cases (% of State Population) for all 50 states(Log-Y)[data via @JohnsHopkins and @WHO March 11 2020] #CoVID19 1/?](https://pbs.twimg.com/media/ES_-PgUXsAA7aKv.png "Total Confirmed Coronavirus Cases (% of State Population) for all 50 states(Log-Y)[data via @JohnsHopkins and @WHO March 11 2020] #CoVID19 1/?")
(Log-Y)
[data via @JohnsHopkins and @WHO March 11 2020]
#CoVID19
1/?
Doubling time (on this VERY noisy and small-n #CoVID19 dataset) appears to be ~3–4 days for most states, with a few outliers. These numbers are almost definitely wrong (and pessimistic), but I'm curious how test-kit availability will affect the #'s over the next few days.
2/?
2/?
Here, y (linear) is total count of confirmed #coronavirus cases per state. Legend has doubling-time estimates for each state, which (as I mentioned upthread) is very noisy, but seems to agree w early Chinese reports.
Yes my coping mechanism is science+data, why do you ask?
3/?
Yes my coping mechanism is science+data, why do you ask?
3/?
Here's estimated #coronavirus doubling-time (Td) as more estimates came in over the past few days.
You can see that we're sorta-ish-kinda-ish converging around 2.5–5 day doubling times. Td=2.5 is really scary. Td=5 is only *mostly* scary. (Spikes on left side are noise.)
4/?
You can see that we're sorta-ish-kinda-ish converging around 2.5–5 day doubling times. Td=2.5 is really scary. Td=5 is only *mostly* scary. (Spikes on left side are noise.)
4/?
Not a particularly attractive graphic, but it's illustrative of why we are way further into this pandemic than it may feel:
Log y (update of tweet #1 in this thread, data via @JohnsHopkins)
5/?
Log y (update of tweet #1 in this thread, data via @JohnsHopkins)
5/?
Things in the US are finally starting to look like the scary exponential #coronavirus curves we've seen elsewhere around the world. Washington's Td is likely only so optimistic bc the plateau artifact messes with curve-fitting on such small n.
[data via @JohnsHopkins]
6/?
![Things in the US are finally starting to look like the scary exponential #coronavirus curves we've seen elsewhere around the world. Washington's Td is likely only so optimistic bc the plateau artifact messes with curve-fitting on such small n.[data via @JohnsHopkins]6/?](https://pbs.twimg.com/media/ETWW2e7WAAAa8UO.jpg "Things in the US are finally starting to look like the scary exponential #coronavirus curves we've seen elsewhere around the world. Washington's Td is likely only so optimistic bc the plateau artifact messes with curve-fitting on such small n.[data via @JohnsHopkins]6/?")
[data via @JohnsHopkins]
6/?
...hooooo-leeeeeey smokes, New York [red line]. Go home and stay home.
Relatedly: What's going on in Washington [blue line]? Is this still a byproduct of data artifacts, I wonder?
#coronavirus
7/?
![...hooooo-leeeeeey smokes, New York [red line]. Go home and stay home.Relatedly: What's going on in Washington [blue line]? Is this still a byproduct of data artifacts, I wonder? #coronavirus7/?](https://pbs.twimg.com/media/ETb_8-0XsAECkRu.jpg "...hooooo-leeeeeey smokes, New York [red line]. Go home and stay home.Relatedly: What's going on in Washington [blue line]? Is this still a byproduct of data artifacts, I wonder? #coronavirus7/?")
Relatedly: What's going on in Washington [blue line]? Is this still a byproduct of data artifacts, I wonder?
#coronavirus
7/?
re: prev img, this is not a prediction — I'm not an epidemiologist — but your HS alg teacher wants me to say:
The #coronavirus Td (y-axis) for NY is similar to other states'. In other words, NY is not worse off. It's just further along the curve, & others are close behind.
8/?
The #coronavirus Td (y-axis) for NY is similar to other states'. In other words, NY is not worse off. It's just further along the curve, & others are close behind.
8/?
I guess this is why we plot things on a log scale!
(Note that the y axes mean different things in these two images)
#coronavirus
9/?
(Note that the y axes mean different things in these two images)
#coronavirus
9/?
Small dataset, but NY's doubling-time is decreasing (≈[disease spreading faster]), while other states' appear to be more-or-less stationary.
10/?
![Small dataset, but NY's doubling-time is decreasing (≈[disease spreading faster]), while other states' appear to be more-or-less stationary.10/?](https://pbs.twimg.com/media/ETmbyURXYAogqCi.jpg "Small dataset, but NY's doubling-time is decreasing (≈[disease spreading faster]), while other states' appear to be more-or-less stationary.10/?")
![Small dataset, but NY's doubling-time is decreasing (≈[disease spreading faster]), while other states' appear to be more-or-less stationary.10/?](https://pbs.twimg.com/media/ETmb1BpWoAM9nlz.jpg "Small dataset, but NY's doubling-time is decreasing (≈[disease spreading faster]), while other states' appear to be more-or-less stationary.10/?")
10/?
As predicted, the percentage of positive vs total #coronavirus tests is really hard to parse visually (cc @huvegi), especially because the number of tests is so low. (If you recently disbanded a US pandemic response office, I blame you.)
...thought continued in next tweet
11/?
...thought continued in next tweet
11/?
defo not the prettiest fig I've ever made, but it helps tell the story. The TOP of the area is the total number of cases. The BOTTOM of the area is # of positive cases.
Most people aren't testing positive, but the ratios differ; i.e. illustrates statewise testing bias.
12/?
Most people aren't testing positive, but the ratios differ; i.e. illustrates statewise testing bias.
12/?
It's been a few tweets since a caveat, so... CAVEAT: I'm still not an #coronavirus epidemiologist. I am Carla and Molly making mozzarella and you are a very patient Italian man politely throwing away all of my garbage work.
13/?
13/?
If the situation in NY vs other states looks like a huge gap to you ("they must have done something differently/wrong/been under different circumstances"), it's because you're not thinking with exponentials. These are the same figure with different y scale:
#coronavirus
14/?
#coronavirus
14/?
It's frustrating to see that these #coronavirus plots are barely deviating day-to-day (such is the nature of an exponential curve; 1000 people changing their behavior today barely looks like a blip, where it'd have been huge last week).
15/?
15/?
What is this?! Au contraire! When plotted on a log-y axis, it becomes clear that the #coronavirus growth curves are bowing under the changes we are making to our everyday behavior.
• Keep washing your hands
• Stay in your house
• Do not follow medical advice from the president
• Keep washing your hands
• Stay in your house
• Do not follow medical advice from the president
When you're feeling bored and stir-crazy and miss your friends, look at these lines slowly curving back toward health. That's why we're staying in!
Been a while since I added to this thread, but... Keep your eyes on the prize. This curve back to normal is a HUGE impact we've made against #coronavirus already:
