Covid-19: Why Aren't We Randomly Sampling?
Individuals want to be tested for Covid-19 but only a limited number of testing kits are available. We can't test everyone.
Shouldn't government officials randomly sample populations rather than testing whomever volunteers? What use is testing if we don't stick to proper statistical sampling methodologies? How can we know how good or bad the situation is if we don't have decent statistics? We _do_ want to know how good or bad the situation is, don't we?
(
5 comments
| 2 points
| on HN
)
Governments are prioritizing tests for people with symptoms or close encounters with confirmed infected people if they're in a high-risk pool.
> How can we know how good or bad the situation is if we don't have decent statistics? We _do_ want to know how good or bad the situation is, don't we?
Tests on the ground don't give a fuck about the macro (and I'd argue they probably shouldn't). When you have a micro-crisis (such as a suspected case in a nursing home), it's very hard to argue for random sampling for the greater good of "better data".
The U.S. will eventually need to go to random testing(maybe 10% of the population) to get a handle on the pandemic but it's a hard test to administer, it also needs to be done regularly, and it costs over $50 per test, so widespread testing will be hard.
Give it 6 months and I can imagine corporate COVID testing labs for most large companies coming online. Damn this is a weird timeline.
There are other ways to cleverly increase test capacities for "explorative" testing by e.g. combining samples from individuals, testing them as a whole and performing binary search on the positive results.
Germany plans to roll out antibody tests in May to find out who is already immune, building up the test capacity for that seems to be a complicated process though.
We cannot make good judgement without decent statistics.
> We _do_ want to know how good or bad the situation is, don't we?
Define "we". I certainly do. Politicians and media have other priorities than I do.