Hacker News Reader

Comment by srinivgp | Hacker News Reader

Absolutely not a distinction without a difference. "Less than 10%" implies, in the vernacular, that they didn't have confidence stating another round bound. That almost certainly it is greater than 1%, probably greater than 2%, and there's a good chance it's greater than 5%. That's how folks read numbers like this. And then they will correctly conclude they should not listen to the CDC because the CDC's numbers seem made up.

If you need more nuance to avoid miscommunicating but you do not give it, you have miscommunicated.

View on HN

Why can't they just give the number and the confidence interval and the methodology they used to arrive at the numbers??

What percentage of the American public understand what a confidence interval is? I have 95% confidence that this number is under 0.1% using the methodology of pulling numbers out of my butt.

I have been in college level stat classes where many people didn't truly grasp confidence interval. While I would love the data, I think it's not really the most important thing during a crisis.

They don't have to say its a confidence interval. Just say something like: "Though outdoor infections are uncommon, how uncommon is uncertain. Based on the limited data available, we estimate outdoor infections to make up anywhere between 0.1% and 10% of total infections". That is honest, transparent, and doesn't require understanding any statistical theory like confidence intervals.

No - but a lot of people by now work with someone or know someone who does understand stats. So people could talk about these things and start to understand it.

If the larger public could read and understanding the underlying data in the way you're proposing then the CDC's guidance isn't of importance. Since they won't/can't, the public depend on/trust the CDC's guidance, who were intentionally conservative here for that exact reason.

As far as public health policy goes this is a distinction without a difference. Everything else is just pointless pedantry.

I can read the data. However I also know it is messy (as all data is), and needs a lot of understanding. Give me the data and some time with a computer and I can make sense of it though - I have all the skills. However I also have a day job, and then a family to spend time with. There isn't enough time left to figure out the raw data. Thus I really need trustworthy people to figure it out for me and give me the summary that I can use. Such people should ideally have practice at the job and so can do it faster than me, while also looking for things that I would forget to account for.

That's exactly what they did.

They said it is less than 10%, and now we have a bunch of people "correcting" them that it could be actually less than 1%. That correction is unhelpful/pointless/pedantic.

If the statement is still true, and the public health guidance is identical, then what is it we're even discussing? This discussion is pointless and the article itself is pointless.

The CDC cannot win no matter what they do. They say it is "less than 1%" then someone will find a study that shows that they're wrong, if they say "less than 10%" then people will say that isn't precise enough and that they're wrong.

> The CDC cannot win no matter what they do.

Their job is to pick the most accurate number they can.

And by accurate I don't mean "most mathematically correct", because that would be "less than 100%".

Making an accurate estimate isn't about winning or losing. It will never be perfect. Oh well, still have to try.

You see how "less than 100%" would be a horrible number to pick, right? Wanting more accuracy is not pedantry.

They probably can't actually know the answer any better than that. Think for a moment about this statement from the article: "There is not a single documented Covid infection anywhere in the world from casual outdoor interactions, such as walking past someone on a street or eating at a nearby table." That's probably true, but how would you even find those infections? There's no record of who walked past who on the street, no way to contract trace every stranger that passes by an infected person, and generally countries don't even try. So the fact that this form of infection hasn't been documented to happen says very little when it's so much harder to detect than indoor transmission. Ultimately, where you draw the numerical line is going to mostly be a matter of personal judgement, and no matter what choice the CDC make the New York Times will always be able to find experts who back a different conclusion.

(Also, the word "casual" is doing a lot of work in that sentence. There absolutely have been documented examples of outdoor transmission via, for example, people going for walks together or chatting with each other outside. So if people think that this means all outdoor interaction is safe, as the article seems to be inviting them to do, this could cause some real problems.)

In your first paragraph, you say there's no way to contact trace outdoors, and in the second you say there are documented cases of it happening.

There are ways to predict your chances of getting infected from someone outdoors. You back-trace from known cases, identify the people they have interacted with, and from those that later contracted Covid, you note whether the contact was indoors or outdoors. There is plenty of data to do that with.

The instances of outdoor infection that have been documented aren't just random strangers passing each other on the street or sitting near each other - as I recall they involved people who knew each other, which obviously makes contact tracing easier. Also, one of the ones I know of was in New Zealand during the period where every case was headline news and aggressively investigated, and the other was in China which similarly had very few cases that were very aggressively traced. This does not say good things about how well countries with more cases are detecting outdoor transmission.

> That's probably true, but how would you even find those infections?

Well, you could look at the infection rates among people who attended large outdoor gatherings during the pandemic in spite of warnings - like Trump rally attendees or BLM protesters.

Or you could look at infection rates for people who have a tractable list of possible exposures - interview only people who live alone and work from home, and see if people who only shop for food have a lower infection rate than people who shop for food and also golf once a week.

Or you could run an experiment - get some infected people, have them cough at petri dishes inside and outside, with and without masks, and at different distances.

Reply on news.ycombinator.com