Read on Twitter
This viral post is correct that you need to think about COVID in the US as a bunch of regional pandemics; but it's not Simpson's paradox. https://twitter.com/mbeckett/status/1278750652160634880
Simpson's paradox would be if deaths per case were rising in every individual region, but paradoxically decreasing in the national total.
If mounting deaths in some places are countered in national totals by an negative trend elsewhere, that's important, but it's not Simpson.
Wiki explanation: https://en.wikipedia.org/wiki/Simpson%27s_paradox
The thread also suggests you'd see deaths blowing up along with case numbers if you zero in on hotspots https://twitter.com/mbeckett/status/1278750656438784001
But it's not clear to me that's true? He suggests looking at Miami and Houston. Here's deaths/day in Harris Co. TX
And Miami-Dade FL (these are both from https://covid19-projections.com/ )
That's FOR NOW and could change. And serious, long-lasting non-fatal illness is a real thing we'd better not ignore.
And deaths in Arizona definitely are up, a lot, though not as much as cases are up.
Sum-up: I don't think the phenomenon of cases increasing while deaths decrease, or increase more slowly, is a statistical illusion.
