Hacker News Reader

Comment by throwaway4good | Hacker News Reader

From the linked article: "Of the 24,079 study participants with blood specimens from January 2 to March 18, 2020, 9 were seropositive, 7 of whom were seropositive prior to the first confirmed case in the states of Illinois, Massachusetts, Wisconsin, Pennsylvania, and Mississippi."

So 9 out of 24,079. Assuming the blood sampling is representative (which it is likely not) and just multiplying up to all 330 mio Americans. It would mean that 123,000 Americans would have had or had the virus at that point.

That sounds like a lot. (And it also sounds it would have been spreading for a while.)

View on HN

> Assuming the blood sampling is representative (which it is likely not)

That assumption is, uhm, doing a lot of work there

Yeah, this is selecting for people participating in a medical study, it is far from a random sampling.

Lots of very rural America would be ruled out, even if the study participants weren't all from traditional big cities.

People that are less mobile also wouldn't participate, etc. Lots of people who wouldn't contract COVID early would also not be likely to participate in this kind of trial.

Plus the positives in this case weren't all in January, they spread into March (by which point in time finding positives becomes much more likely).

And that's aside from the difficulties with antibody testing and cross-reactivity -- although this study looks a lot better than the previous one.

Having worked with banked blood sample data before for influenza research, it's very, very much not representative.

> Assuming the blood sampling is representative (which it is likely not)

I can't think of anything more anti-science than dismissing data.

how is he dismissing data? questionable data should be treated as such... why would a linear extrapolation be a good fit here?

I think they are implying that blood sampling is not a random sample of people (more likely to be from someone who is sick), not that it is inaccurate.

I don't think the phrase you quote is intended as a dismissal of the data. It comes across as an acknowledgement of the data's limitations, and a caution about what kind of conclusions can be drawn. During my statistics course in college, one of the first sections of the course covered sampling, and the various ways sampling can be done badly, thereby biasing the sample, and thereby leading to conclusions about the sampled population that don't reflect the reality of that population.

Cautioning against that failure mode strikes me as intellectually honest, not anti-science.

Reply on news.ycombinator.com