This tweet and the associated thread is such a perfect variation of the famous statistical fallacy. Silver looks at a change in cases and relates them to the initial level, inserts some semi-causal views about why the levels vary, and then declares that there is a group in the middle that's "stuck." Once average, always average.One observation related to this:
— Nate Silver (@NateSilver538) May 6, 2020
States that have seen clear declines in new cases tend *either* to have had either a LOT of cases early on (e.g. NY, LA) or very FEW cases (e.g. MT, HI). The states in the middle (e.g. MD/VA) aren't seeing much, if any, decline, conversely. https://t.co/NhLUwFu0sF
The problem is randomness. Random things (e.g. superspreaders, accidents, things we still don't know) will move states around these categories over time. And with each of those moves, people will come along with new ex post rationalizations for why a particular state is low or high -- and again to seek explain a group "stuck in the middle." It's also the road to leads to inventing epidemiological terms like "partial herd immunity."
It's easy to be fooled by randomness when you forget it's there.
No comments:
Post a Comment