Thread by @midnucas

One thing I keep saying is that when you see whopping-high positive rates like this one you should pause and check if they're true. I checked this one against South Dakota's website, which says 20.3%.

But I think BOTH are wrong.

https://twitter.com/JesseRodriguez/status/1327036402543042572

Using @COVID19Tracking data, I can reproduce the 68% calculation. It's the Nov. 12 cases figure divided by the new PCR tests (people) figure: 2,020 ÷ 2,968 = 68%.
covidtracking.com/data/#state-sd

But:

• The cases figure includes probable cases which are detected through antigen tests, not PCR, so don't figure in the denominator.

• Don't frickin' calculate positive rates for individual dates, take 7-day averages.

So I get:

8,605 confirmed cases (7 days)
÷ 16,952 PCR tests (people, 7 days)
───────────────────────
56.4% positive rate (PCR, cases ÷ people, 7 days)

So how come South Dakota says getting ~20%? I think they're dividing people ÷ specimens 🤦‍♂️:

8,605 confirmed cases (7 days)
÷ 42,800 PCR tests (specimens, 7 days)
─────────────────────────
20.1% positive rate (PCR, cases ÷ specimens, 7 days)

The COVID Tracking Project has an article explaining what "specimens" are, but in short, according to South Dakota's data, each test subject got swabbed on average 42,800 ÷ 16,952 = 2.5 times over the last week.

"For our part, The COVID Tracking Project does not directly calculate test positivity rates—and we will not do so until we are confident in our ability to communicate precisely about these complex issues in our annotations and visualizations."
covidtracking.com/blog/test-posi…

Forgot to link the South Dakota Department of Health coronavirus page: doh.sd.gov/news/coronavir…

I should add: Don’t trust my 54% calculation above either, because I don’t know the mechanics of how South Dakota’s deduplicated test numbers work. If I got a test a month ago, do I make the count rise if I got another last week? Mega caveat emptor.

https://twitter.com/midnucas/status/1327052007652225024