What regression to the mean is: The phenomenon in which a variable is extreme upon the first measure, but moves towards the mean, or average, through subsequent measures. Having a coin land heads up 10 times in a row, for example, represents an extreme but not impossible event. However, if the coin is fair, flipping it 10 or 20 more times will produce a result more closely aligned with the standard probability of landing heads up.
What regression to the mean is not: A causal phenomenon. Independent events do not “even out” or “compensate” for extreme observations. Rather, the regression to the mean simply describes a tendency for probabilities to normalize as samples increase.
Why regression to the mean matters: Accounting for regression to the mean is critically important when designing experiments and interpreting data. Without doing so, incorrect inferences, commonly called regression fallacies, are likely. When designing a test, it is important to collect enough data across a representative time period—often, more and longer than a basic significance test may require—to establish confidence in a result.
Enjoy this post? Download all 200+ terms!