4 ms·
CDC: Coronavirus antibody tests may be wrong 50% of the time
- tinus_hn 6y agoNo, that is not what they say. They say that because the number of people that are actually positive is low, the number of false positives may be high when compared to the actual number of positives. This does not mean it’s wrong 50% of the time, it’s a statistical trick.
- im3w1l 6y agoIn jargon, the precision is lower than 50%. https://en.wikipedia.org/wiki/Precision_and_recall https://en.wikipedia.org/wiki/Precision_and_recall
- contravariant 6y agoTechnically that description is correct, but people might confuse that definition of precision with the false positive rate, which usually does not include the prior probabilities.
- tomohawk 6y agoCDC quote: > Serologic test results should not be used to make decisions about returning persons to the workplace.
- m0llusk 6y agoRight, this is more about tracking clusters and finding out exactly how effectiveness of different regimes such as lockdowns and required mask wearing compare. We need all the data we can get to make sure that our social policies are correct based on robust scientific analysis and not on emotions conjured from the demon haunted world.
- dbbk 6y agoAre they referring to antibody tests in general as a concept, or a specific test?
- taeric 6y agoThis wording is confusing. With our priors of how many real positives there are, a positive result could be wrong fifty percent of the time. The same is not necessarily true for negatives. If our priors are wrong, of course, the test could be wrong more or less. Which is to say, yes, it is complicated. Edit: and I should add that my understanding here could be wrong. Would appreciate elaboration on the point.
- m0llusk 6y agoThis article is woefully vague. There are multiple competing antibody tests that use different production methods. The Abbot Labs antibody test appears to exceed 99% accuracy which makes it significantly less susceptible to this problem. Additionally, the common expectation is that antibody tests should be quick and cheap and easy, possibly easy enough for people to perform themselves at home. This makes it possible to generate enough additional test data to correct for some inaccuracies.