Regression to the mean
Also called: Mistaking a bounce for a cure
Crediting a treatment for a change that would have happened anyway, since extreme results tend to return to normal.
How to spot it
Was the change measured after an unusually bad or good time?
Example
“I only called the doctor on my worst day, and I felt better the next day. The call helped.”
How to answer it
Ask what would have happened with no treatment at all.
Related
- Post hoc: Assuming that because one thing followed another, the first caused the second.
- Correlation is not causation: Assuming that two things which move together must have a cause-and-effect link.