Read on Twitter
San Diego here is demonstrating a troubling attribution / statistical interpretation error I've seen more and more -- people are declaring "success" of their homelessness initiatives based on comparing rates inappropriately. It's a bit wonky but here's what I mean... (1/) https://twitter.com/LisaHalverstadt/status/1357174962218442756
Here San Diego reports a 2.3% positivity rate among homeless people observed from "mass testing of asymptomatic residents." They report this compares favorably to an 11.9% positivity rate observed in region-wide testing. (2/)
But testing for #COVID19 in a community overall is not universal or random — many people seek testing because they have symptoms or known exposures — making the denominators for the two groups quite different and thus these rates aren’t directly comparable. (3/)
You would expect positivity rates for universal or surveillance-type testing of asymptomatic people to be significantly lower than rates for testing for a group obtaining tests for a variety of reasons including, in large part, tests prompted by symptoms or exposures. (4/)
Of course some people in the community get tests when they are asymptomatic or have no risks but still the denominators for these groups are different enough that it's akin to comparing apples and oranges...and in a way not easy to precisely quantify. (5/)
San Diego isn’t the only city to have made this error in comparisons. I’m not sure if it's due to officials misunderstanding concepts, willful obfuscation (lies, damn lies, and statistics?), or if something is lost in translation to the public. I suspect some combo of 1 & 3. (6/)
Hope this all makes sense. What then would be a better comparison? You'd have to construct a comparison group with testing done more similarly -- such as a random community sample including largely asymptomatic people. (end)
