Big healthcare Q1 losses and layoffs in Detroit.
Printable View
The arguments don't explain away the cases. They explain away confidence.
The false positive rate may actually be tiny, but the internal and manufacturer testing was phenomenally unlucky and had a bad run of false positives. In that case, the actual number of seropositive people could be even greater than 50-85 x the number of cases.
The arguments show how we can't really draw any conclusions either way with even a moderate degree of certainty.
For some reason, when I include the quote of FlamingoDeFranc's post about the two arguments canceling out, my post doesn't go through.
Here's my response:
The two arguments address separate possible scenarios, and thus cover more ground together. They complement each other.
Here's an analogy:
Suppose there's a test to detect injuries due to barbell training, and suppose someone runs a study to measure barbell injuries and reports that 15% of their tested population tested positive for barbell injuries (and then go on to make the argument that barbell training is dangerous).
Here are two arguments that show flaws with the study:
1) The test has false positive rate that could well be 15%. So it could well be the case that there were actually 0 individuals with barbell injuries, and that the 15% they found were all false positives.
2) The tested population was wholly recruited from a local rehab clinic, where individuals would be more likely to have injuries. Thus, even if all 15% of them actually did have barbell injuries, we
cannot generalize this number to the general population.
These two arguments are structurally identical to the ones we're discussing. They certainly do not cancel each other out. They form a stronger case against the study together than alone.
"Its Under the three scenarios for test performance characteristics, the population prevalence of COVID-19 in Santa Clara ranged from 2.49% (95CI 1.80-3.17%) to 4.16% (2.58-5.70%). These prevalence estimates represent a range between 48,000 and 81,000 people infected in Santa Clara County by early April, 50-85-fold more than the number of confirmed cases."
If I'm reading this correctly, they are saying that the amount of people infected (non-cases + confirmed-cases) is 50-85 times more than the number of confirmed cases alone. They don't mention fatalities in the abstract at all, as that's not what the study was designed to do. They simply want to know how many people actually have the disease. If the criticisms of the paper hold weight, that would mean that the prevalence of the virus alone (never mind cases, deaths, recovery, etc) was way over blown, meaning that the disease is not as infectious as they initially concluded. How much less infectious? Who knows. Probably not THAT much less.
Yes, I meant infections. Affects the IFR, not the CFR.
sorry, i take the bolded part back. It should read:
The false positive rate may actually be tiny, but the internal and manufacturer testing was phenomenally unlucky and had a bad run of false positives. And if you combine that possibility with the possibility that the test also happened to have a phenomenally bad sensitivity that went unnoticed because of an extremely lucky streak during validation, the actual number of seropositive people would be ever higher than their estimate.
As Gov. J.B. Pritzker says COVID-19 won’t peak in Illinois until mid-May, Mayor Lori Lightfoot predicts statewide stay-at-home order could extend into June - Chicago Tribune
Can someone explain to this guy that saving a few thousand lives isn't worth ruining millions? Or does that hurt his chances of reelection?
Identify the mathematical statistician among the authors. The one who computed Var(s) in their appendix.
Biased samples aren't good, but can be corrected via poststratification. Mathematical errors, on the other hand, kill the paper.