by rishirishi on 4/26/20, 8:46 PM with 2 comments
by joshdance on 4/27/20, 4:05 PM
Especially when what they're doing isn't actually extrapolation, it's statistical inference using a sample that's not random.
Context: I just watched a video where an urgent care doctor took the % of people testing positive, multiplied it by the total state population, inferred the total number infected across the state, divided the number of deaths by this, & then told people their odds of dying.
You can't say that 12% of the entire population has contracted COVID-19 just because 12% of the people tested are positive if the vast majority of those being tested have symptoms. Healthy people are underrepresented in that sample.
That would be like trying to estimate the percent of the entire population that's vegetarian by only surveying people at an In-n-Out burger joint. I mean the fries are great and everything, but…”