Originally Posted by
Shiva Kaul
The first issue is: their dataset has only 50/3300 positives. It is possible (within the 95% CI) that all of those are false positives, i.e. there is zero seroprevalence in Santa Clara. I identified the reason why zero seroprevalence escapes their reported CIs - basically, a dubious modeling assumption.
The second issue is a seeming mathematical error, which also casts doubt on their CIs.
Unlike others, I’m not out for the authors’ blood. They can fix this. But, as is, I don’t think those 50-85x conclusions are justified, and the whole thing could be a wash.