Marin in 24 hours. 1,530 tests completed. 8 new positive cases. 1 death.
GIGO? Garbage in / garbage out.
It's a common expression in analysis. Your model may be great, but if the data are no good, your output's going to be bad.
An example. The SFChron reports on a model from Georgia Tech (good school!) that tries to predict what the probability is that someone in a crowd of 25 people has COVID. From the article (and note, the Marin projection is badly wrong): "A quick browse through the map yields some troubling numbers. If you’re in San Francisco County and decide to attend a gathering of 25 people, there’s a 34% chance that someone who’s coronavirus-positive will be in attendance. In Marin County, the risk is 75%. In Alameda County, it’s 31%. Increase the gathering size to 50, and the risk goes up significantly. In San Francisco County, the likelihood is 56%; in Marin County, it’s 94%."
Two errors apply to the Marin case.
First, and worst: they include the San Quentin outbreak in the Marin general population. That's 2,212 cases on top of the nearly 3,000 cases in Marin's general population: a major difference.
Second, and more subtle: the estimate of seroprevalence is 10:1 x the known count. They acknowledge that this estimate will be too high in areas where there is a lot of testing (like Marin, where completed tests exceed 53,000 - 21% of the county's population). The online tool allows adjustment of the seroprevalence ratio to 5:1. I did that and the risk of someone in a group of 25 in Marin having COVID dropped to 31%. That still includes the SQ cases in the general population.
The risk in Marin - especially the south of the county is closer to and likely lower than that in SF. I'm not saying it's safe to gather in crowds. Wear masks! Keep distance! I'm saying: the SF Chron article is wrong. (I've written to Georgia Tech and offered to help them correct their data.)
Map attached from Georgia Tech: the whole USA map is their general case w/ seroprevalence ratio at 10:1. The zoomed-in map of Northern California has the ratio set to 5:1.
Maps attached from Georgia Tech. Model at https://covid19risk.biosci.gatech.edu/, Chronicle article at https://www.sfchronicle.com/bayarea/article/If-25-people-gather-in-SF-odds-are-34-that-at-15458554.php
