Regression to the Mean

Regression to the mean is a statistical phenomenon in which extreme or unusual results tend to move closer to the normal result on subsequent measurements. In other words, if something measures unusually high or low on one measurement, it is likely to be closer to the mean on the next one.

For instance, imagine a student who takes weekly math tests typically gets 70%. One week, she gets 90%. The next week, it would not be surprising if her score was somewhere around 70% again. This is regression to the mean.

Not recognizing the possibility of regression to the mean can lead to attributing causality when it's not there.

For example, imagine the teacher had praised the student when she got 90%, and the next week, when the student got 70%, the teacher thought, "Praise made her perform more poorly." This teacher neglected to consider that the lower score was likely just a case of regression to the mean.

As another example, if a company implements a new training program after a particularly bad sales month and then sees sales improve the next month, they might attribute the improvement to the training. However, sales were likely to have gone up anyway due to regression to the mean.

Understanding regression to the mean is important for data analysis because it helps prevent misattributing changes to interventions that are normal statistical fluctuations.

Back to the Data Analysis Handbook