FILE SELECTED // FILE 00
Regression to the Mean
“Cases selected for being extreme often look less extreme next time, even without an intervention.”
01 / SUMMARY
In short
When a measurement combines a fairly stable signal with random variation, selecting extremes favors observations that also had extreme noise. On repetition, that noise is usually less extreme.
03 / SCENE
In the wild
People are rewarded after their worst week and many improve; part of the change would have happened without the reward.
04 / LIMIT
The limit
It is not a force pushing individuals toward average, nor does it guarantee every case will move closer.
FINAL NOTE
What remains
When improvement starts from an extreme, compare against a group or several prior measurements.