re: Stanford University anti-body study finds COVID-19 more widespread than thought

quote:

---

Data from telluride testing:

---

And San Miguel county has 13 positive cases (0.16%). Your article is a bit outdated, but currently 17 of 2583 have tested positive 0.66%, which equates to 54 cases county wide, which means 4.2 times more cases than currently identify.

Extrapolating that nationally, that equates 2,949,318 so with 37,158 deaths, that’s an IFR of 1.26%.

quote:

---

Well 60% asymptomatic (although some could be pre-symptomatic) means that if we’re only catching symptomatic cases, that there are 2.5 times more cases than those identified. And if we’re only catching 50% of the symptomatic cases, and double 2.5, that puts our current IFR at 1.05%.

---

Half the people with cold-like symptoms are going to the hospital and getting tested for covid? Nonsense.

quote:

---

And that's a nice attempt at mental gymnastics, but how about we go with
1/660 = 0.15%

---

So the group that is disproportionately less likely to die than pretty much any other group, as a result of the combination of their age and physical health, and we’re looking at a fatality rate that’s at best 50% higher than the seasonal flu.

quote:

---

Nobody else is expected to die.

---

There is literally nobody who could possibly know this, even their own physicians, let alone some random person that doesn’t know anything about any of the remaining 659 individuals.

Regardless, a 0.15% IFR given that group, means that the population’s IFR is far higher.

quote:

---

Half the people with cold-like symptoms are going to the hospital and getting tested for covid? Nonsense.

---

You think the symptomatic cases are just “cold like” symptoms?

Regardless, people were lining up to get the tests when they are allowing pretty much anyone to take it (before restricting it again to those more likely to be positive), and this study filled up with participants almost immediately.

So I would say that those who were symptomatic were more likely than not to at least seek out getting tested.

You’re reporting specificity of the tests: 369/371 = 99.5 and 368/371 = 99.2. Where are you getting 366? I didn’t see that anywhere in the link you provided. Also, why are you specificity to conclude that false positives are 45 out of 50. Specificity measures the fraction of those who are truly negative.

You seem to be arguing the positive predictive value is actually off, i.e. instead of their being 50 true positives there are only 5, hence 5/50 = 0.10 or 10% of those positives were true positives. If that were true, then the specificity numbers would be way off.

Did the paper show the numbers of false +/- and true +/-? I saw the sensitivity and specificity numbers but not all of them. Perhaps, I can backdoor it.

quote:

---

Where are you getting 366?

---

If EITHER the LGG and/or LGM test is positive, then it’s considered a positive test. Therefore, a negative test is when BOTH are negative.

So while it doesn’t proved the classification accuracy when both are negative, we can estimate it by the accuracy of the LGG times the accuracy of the LGM (.9946*.9919=.9866, which is 366/371).

quote:

---

You seem to be arguing the positive predictive value is actually off, i.e. instead of their being 50 true positives there are only 5, hence 5/50 = 0.10 or 10% of those positives were true positives.

---

That’s exactly what I’m saying.

quote:

---

If that were true, then the specificity numbers would be way off.

---

Well using the 98.66% specificity, we get 45 false positives with a 100% negative sample. And we also get 45 false positives with (0.15% positive) and 5 true positives which results in 50 total positives.

So given these data, I think that would be the most best crude estimate, even though it’s possible all 50 were false positives.

quote:

---

Did the paper show the numbers of false +/- and true +/-?

---

They used the BEST single antibody test estimate from the manufacturer (LGG test; 369/371) which is a bit deceiving since they are using both tests and the LGM is lower (368/371) and the maximum upper bound in this case.

In addition, they did their own control analysis, with a sample of only 30, which is pointless since even a test with 97.75% accuracy has a better than 50% chance of getting 100% correct with a sample size of 30.

And as I’ve pointed out before, this test is an unapproved and test from China that has now been banned by China for exporting. The first FDA approved test, that uses the same methodology, had a 96% specificity. So I feel like 98.66% is generous.

quote:

---

There is literally nobody who could possibly know this, even their own physicians, let alone some random person that doesn’t know anything about any of the remaining 659 individuals.

---

Actually, you were wrong. A simple report of the number of individuals who were sick enough to go to the hospital (publicly available) was all you needed.