This is simple textbook regression to the mean fallacy. With randomness in the data, high observations are more likely -- purely for probabilistic reasons -- to be followed by lower ones, and low observations to be followed by high ones.That's part of why the national numbers have looked pretty good even though there are a lot of states to be worried about. The numbers tend to be either high but falling, or low but rising—but not many places where they're high and rising.
— Nate Silver (@NateSilver538) June 5, 2020
It should be embarrassing that such a basic statistical error keeps being repeated, and to a large audience.
No comments:
Post a Comment