Apropos of this point and for my own edification, I& #39;m going to use this thread to collect links to stories or other information about how the coronavirus-data sausage gets made, and how that shapes what we see in it. https://twitter.com/NateSilver538/status/1262089315695329283">https://twitter.com/NateSilve...
China& #39;s coronavirus toll is surprisingly low. This @100Reporters piece from a week ago discusses a leaked database which implies that China has had an order of magnitude more COVID cases than its official statistics show. https://100r.org/2020/05/china-coronavirus/">https://100r.org/2020/05/c...
From @maggiekb1 at @FiveThirtyEight, a great and thorough look at how cause of death is determined, and why this implies that COVID deaths in the U.S. are almost certainly undercounted in real time. https://fivethirtyeight.com/features/coronavirus-deaths/">https://fivethirtyeight.com/features/...
. @AP on the pandemic& #39;s toll in Manaus, Brazil, where deaths are running 3 times as high as usual, but "due to a lack of testing, just 5% of the more than 4,300 burials performed in April and May were confirmed cases of COVID-19." https://apnews.com/54423b73d8be5fbc2a491bcc60f10285">https://apnews.com/54423b73d...
One takeaway from all these pieces is that, at present, broad cross-national comparisons of COVID-19 epidemics are nearly worthless. In many cases, uncertainty around the true toll is so high that even just rank ordering is inappropriate.
Another is that the uncertainty skews hard in one direction. While there are errors of commission (e.g., false-positive tests, mislabeling deaths as COVID-related), they are generally swamped by errors of omission (e.g., lack of testing, political pressure to suppress count).
You can follow @JayUlfelder.
Tip: mention @twtextapp on a Twitter thread with the keyword “unroll” to get a link to it.

Latest Threads Unrolled: