Over on Twitter Taal Levi (@taaltree) posted a long thread explaining statistical issues in COVID-19 testing in great detail. The issue is the relationship among the specificity and sensitivity of the test as well as the actual number of people infected and not infected. Lucas C. Wheeler added a nice graphic representation of the problem. I’ll show you Wheeler’s work first then Levi’s thread as a pdf file.
Lucas C. Wheeler (@LCWheeler9000)
4/10/20, 6:15 PM
@taaltree I tried to make a pair of visualizations that show what what this thread is saying in a way that I hope is easy to understand. First one has incidence = 1% and false/real=82%. Second one has incidence=10% and false/real=30%. Here you go!
(click for larger image)
(click for larger image)
