News for hackers. Stuff that matters.
Re:In 4 US state prisons, 3,300 inmates test positive, 96% w... by raverbashing on Saturday April 25, 2020 @05:38PM (#22981763)
I wonder what's the process through which false positives happen in this case. Previous infection by milder Coronaviruses?

Edit: I'm looking at the reddit post but I have a lot of reservations with the "prevalence 0.06", unless we'll use the test to test absolutely everybody and not only people who are suspect. Has that calculator been validated as well?

If the test was 12 false positives in 128 negatives, how come they can claim the false positive rate is 60%?

Re:In 4 US state prisons, 3,300 inmates test positive, 96% w... by ptrincr on Saturday April 25, 2020 @05:52PM (#22981883)
Apologies for the way this was linked to. The 6% is from this study:

https://www.miamidade.gov/releases/2020-04-24-sample-testing...

"Our data from this week and last tell a very similar story. In both weeks, 6% of participants tested positive for COVID-19 antibodies, which equates to 165,000 Miami-Dade County residents"

That is what the commentator is referring to in the linked post.

So if you plug their own figures into the calculator:

Sensitivity .8866 Specificity .9063

and a Prevalence of .06 based on the study, you get the 62% false positive rate.

As the prevalence increases, as with the NYC study which found the positive rate to be 21% (prevalence), the false positive rate decreases, down to 28% of the NYC study.

Re:In 4 US state prisons, 3,300 inmates test positive, 96% w... by Symmetry on Saturday April 25, 2020 @08:51PM (#22983175)
The password you need to Google for why it happens is "antibody cross reactivity." Not necessarily other coronaviruses but I imagine they're disproportionately more likely to cause it.
Re:In 4 US state prisons, 3,300 inmates test positive, 96% w... by suzzer99 on Sunday April 26, 2020 @03:18AM (#22985246)
This is from ARCPoint Labs, where I took my antibody test:

The Antibody test is a serology test which measures the amount of antibodies or proteins present in the blood when the body is responding to a specific infection. This test hasn’t been reviewed by the FDA. Negative results don’t rule out SARS-CoV-2 infection, particularly in those who have been in contact with the virus. Follow-up testing with a molecular diagnostic lab should be considered to rule out infection in these individuals. Results from antibody testing shouldn’t be used as the sole basis to diagnose or exclude SARS-CoV-2 infection. Positive results may be due to past or present infection with non-SARS-CoV-2 coronavirus strains, such as coronavirus HKU1, NL63, OC43, or 229E.

Re:In 4 US state prisons, 3,300 inmates test positive, 96% w... by FrojoS on Saturday April 25, 2020 @06:00PM (#22981947)
Yes. That’s one possible explanation. Interestingly quite a lot of people might be somewhat immune to the new Corona virus due to anti bodies from previous Corona cold infections. More than 30% showed such antibodies in a recent study. https://www.finanzen.net/nachricht/aktien/drosten-hinweis-au... (Sorry that the only source I have ready right now)
Re:In 4 US state prisons, 3,300 inmates test positive, 96% w... by nodamage on Saturday April 25, 2020 @06:01PM (#22981961)
60% is the probability that a particular positive test result is actually a false positive. It's not the overall the false positive rate.