Thread by @ashishkjha

I don't buy Sturgis study below.

Why? Really big effect. Makes me skeptical

One rule of thumb I teach is:

if effect size of a social phenomenon using noisy data is very large -- be wary

So was Sturgis harmless?

No.

But I doubt it caused 250,000 cases.

Thread

https://twitter.com/SDSUCHEPS/status/1302480031638147074

First, there are a bunch of methodologic issues and two different threads do nice job sorting them out

One by @AssumeNormality

https://twitter.com/AssumeNormality/status/1303427792693059587?s=20

Second by @RexDouglass

https://twitter.com/RexDouglass/status/1303379252742479872

But there's another reason beyond these to be skeptical

2/6

It doesn't pass the sniff test

Yes it was a crazy large rally

Yes there was a ton of irresponsible behavior (no mask, no distancing, lots of bars in evening)

I would not be surprised if it set of big outbreaks -- hundreds of cases, may be thousands

But 250K??

3/7

One way to is to look at where the ralliers came from -- 10% from the Dakotas, 90% from elsewhere.

So let's look at the Dakotas -- actually, here, the story is pretty good.

It appears that in the days following the end of the rally, large spikes in cases in the Dakotas.

4/7

So that checks out

Across Dakotas, likely additional 1K to 2K cases total since rally

Not huge -- but in Dakotas (small populations), you see it

What about other 90% of Ralliers?

Where are they from?

Authors identify them coming from big counties in AZ, CO, CA, NV, etc

5/7

But when you look at those counties, no big (or even small) uptick in cases in days/weeks following Sturgis

Not in Maricopa, SD, LA, Denver or Las Vegas

If none of the big counties saw a spike, 250K additional cases seems implausible.

6/7

If Sturgis really was 20% of all cases in US since rally, we should see counties where people went back to light up.

They don't

These counties rising pre-rally and flatten post rally (mostly).

Synthetic counties may be not good controls

But here's the lesson

7/8

If Sturgis is responsible for 20% of cases in the past month, we should be able to see it in the raw data.

Because that's a massive effect. Massive.

But we don't

If I'm missing it or am wrong, would love to know how

8/9

Final thoughts:

1. Sturgis rally was wholly irresponsible. 500K folks gathered with little respect for COVID

2. LIkely super spreader event, causing thousands of cases, some deaths

3. But think unlikely to be responsible for 260k cases (20% of all US cases)in short order

Fin