[stupac2]

Nate Silver argues that, given the problems with other sources of data, the best thing to look at is percentage of positive tests: https://fivethirtyeight.com/features/cor...-s-states/ (that article is about something else but it's discussed in there, or at least linked).

This isn't all that intuitive, but I think the various modeling scenarios show it to be the case.

[stupac2]

Nate Silver argues that, given the problems with other sources of data, the best thing to look at is percentage of positive tests: https://fivethirtyeight.com/features/cor...-s-states/ (that article is about something else but it's discussed in there, or at least linked).

This isn't all that intuitive, but I think the various modeling scenarios show it to be the case.