Hacker News Reader

Hacker News

Comment by MrPatan | Hacker News Reader

Full thread

MrPatan

Does anybody know the false positive rate of the test?

Could this be an artifact of that plus low actual incidence like:

- False positive rate of 10%

- 100 tests

- 1 true positives. All with symptoms

- 9 false positives. All without symptoms (duh!)

Headline: "90% asymptomatic!"

Truth: "Shitty test procedure!"

Yes, yes, I know you know this. Do you think people doing tests and writing headlines know this?

What is the false positive rate?

View on HN

It's a prison population. Prisons and nursing homes are both seeing shockingly high infection rates because of the nature of what they do, which involves confining groups of people in close quarters.

People in nursing homes are elderly. Lots are dying.

People in prison aren't uniformly elderly. You are bound to see more variation in symptoms.

(Plus, as stated elsewhere, the opinions and experiences of known criminals tend to get discounted, so the report of lack of symptoms may be more about that than about the general resilience of the population. Also also: It's well established that if you ignore, dismiss and neglect someone enough, they stop complaining because they know it doesn't do any good. Aka learned helplessness.)

" Do you think people doing tests know this?"

Probably.

" Do you think people writing headlines know this?"

Maybe, but unlikely that they care, if the result is a flashing headline.

> They started with the Marion Correctional Institution, which houses 2,500 prisoners in north central Ohio, many of them older with pre-existing health conditions. After testing 2,300 inmates for the coronavirus, they were shocked. Of the 2,028 who tested positive, close to 95% had no symptoms.

The article almost seems to be making a point of not putting all the needed numbers together for any given population that they talk about though.

This is the relevant response. What the commenter is trying to imply is that the authors of this study are so staggeringly stupid that they overlooked the possibility of false positives when designing this experiment.

In reality, this test would need a false positive rate of over eighty percent to explain this kind of asymptomatic infection rate.

Also, prisons are useful because due to the close quarters it can be taken as a given that a substantial proportion of the population is infected, further minimizing the danger of these sorts of errors. The choice of population suggests a sophisticated experiment design, and the commenter is implying that the study authors made a statistics 101-level error.

As someone else has said on another comment, it just needs the experimenter to be infected and a bit careless, and there is your high percentage of positive results.

Plus yeah, to be honest, you dismiss staggering stupidity leading to juicy headlines at your own risk.

No, that can’t be the case here, because a shockingly large proportion of people tested positive according to the article. (Thus, either the false positives are negligible, or the test employed has a an astronomical false positive rate (bad specificity)).

You're right, they get a positive result for almost all the population!

Wait, does that pass the smell test? Do prisions become huge flu hotspots as well? The Diamond Princess outbreak didn't have that kind of numbers... What's more likely at this point, the numbers from the article or human error?

My respects to anybody trying to do actual science with this kind of data in this kind of situation...

Densely crowded environment where people were unlikely to maintain social distancing, poor quality health care, probably poor quality sanitation. It seems reasonable to me that a prison would fare worse than a cruise ship.

Not quite this huge, but then again states generally try to get them flu vaccines. It's reasonably common for prisons to enact emergency measures to stop the spread of the flu.

Right, it could literally be as dumb as: all the tests were processed by one person, and that person is infected with covid-19.

Drawing signal out of the noise right now is very, very difficult.

That's it, you nailed it.

What's more likely, a population that gets all infected all at the same time, or one person infected being careless with samples?

Then there's more: given how many tests of similar populations are getting done, what are the chances that some have a careless infected tester?

A cruise ship under quarantine is dramatically more socially isolating than a prison.

You would expect spread to be absurdly more efficient in a prison. Less physical separation, less hygiene, less everything.

Wild speculation is not the solution to lack of information. We all want to know how reliable these tests are. I don't understand why we can't find this information. But that is not an excuse to make nonsense assumptions and craft imaginary narratives.

Reply on news.ycombinator.com