3 ms·
Yes. The tests are sufficiently unreliable that if there is no specific reason to believe that a particular person has been exposed, the results are just noise
by btilly 7y ago
Yes. The tests are sufficiently unreliable that if there is no specific reason to believe that a particular person has been exposed, the results are just noise.
See https://towardsdatascience.com/statistics-and-unreliable-tests-coronavirus-is-difficult-to-contain-e113b5c0967c https://towardsdatascience.com/statistics-and-unreliable-tes... for an analysis using Bayesian statistics that demonstrates why you would pre-filter who you give the test to, and why you shouldn't necessarily trust the test results after.
Please remember that public health organizations have over 150 years of experience with how to make the most effective use of existing resources to limit health challenges. This is a complicated optimization problem and the people doing it aren't stupid. It is better to ask why they do what they do than to assume that what they are doing makes no sense.
- xorfish 7y agoThe problem in such cases is the prior. If a false positive is 10%, 1%, 0.1% or 0.01% makes a huge difference. With 0.01% and two different tests, you will get a false positive in every 100'000'000 tests. Testing twice solves many of the issues. It also gives information about the sensitivity and specificity of a test.
- wahern 7y ago> With 0.01% and two different tests, you will get a false positive in every 100'000'000 tests. That doesn't sound right to me (i.e. only the case if false positive distributions are independent and uncorrelated, which seems unlikely), but I don't know enough about statistics and epidemiology to dispute it.