percentpositive != caserate / testrate
tonetto opened this issue · comments
Hi, in my understanding, the weekly percent positive
(P) should (at least nearly) match the ratio of case rate
(C) and test rate
(T), but it's never the case. In either city, borough, or ZCTA, P > C/T
is always the case.
You define in your documentation:
percent positive := number of positive tests / number of all tests
case rate := number of positive tests / 100 000
test rate := number of all tests / 100 000
So, just sounded intuitive that I could verify the percent positive, but those don't match. Sorry if I am missing something obvious here, and thanks so much for publicly sharing these data!
Hello, thanks for the question.
The reason that percent positive doesn't match case rate / test rate is because the case rate is not the number of positive tests / 100,000. A person may test positive several times but is only diagnosed as a case once - upon their first positive test.
Additionally, your definitions/interpretations are slightly off - percent positive is the percent of people tested who test positive, not the number of positive tests divided by the number of total tests. Similarly, case rate and test rate are the number of people diagnosed and people tested, respectively, per 100,000. That is, all of these metrics are deduplicated by person for the unit of time reported.