平均への回帰

Regression toward the mean is the tendency for extreme values to move closer to the group average upon remeasurement. This statistical phenomenon results from random error and chance. For example, an athlete's exceptional performance is often followed by average performance, or a patient's extremely high blood pressure reading typically decreases on retesting, not due to treatment but because extreme scores are naturally rare.

Regression toward the mean can create the false appearance of treatment success when none exists. If researchers select participants with extreme initial values and observe improvement on retesting, the change may reflect natural statistical regression rather than treatment efficacy. Without accounting for this phenomenon, researchers might publish ineffective interventions as successful, as occurred with some childhood obesity programs.

Descriptive statistics summarize data from a specific sample, such as average birthweight and standard deviation. Inferential statistics draw conclusions about larger populations based on sample results. Regression toward the mean complicates inferential statistics by potentially making random differences appear significant, leading researchers to incorrectly infer population-level treatment effects when differences stem from chance alone.

Extreme outlier results are rare occurrences in normally distributed data. When retested, these values typically regress toward the mean, appearing less extreme simply due to statistical chance. A physician who observes a patient's extremely high blood pressure and retests it may see improvement without any intervention, potentially leading to unnecessary treatment decisions if the phenomenon is not understood.

Regression toward the mean extends across numerous disciplines including finance, psychology, and sports. It has been observed in stock market performance, flight student performance evaluations, and even divorce rates among couples. This wide-reaching statistical phenomenon affects any field using statistics to draw conclusions, making awareness essential for researchers across medicine, science, and social sciences.

Researchers can use inferential statistical methods specifically designed to evaluate and account for regression toward the mean. Strong evidence of treatment efficacy comes when a therapy lowers an entire group's average, not just improves extreme cases. These statistical approaches increase confidence in data veracity and allow researchers to distinguish genuine treatment effects from natural statistical regression.

Retesting reveals whether initial extreme values represent true characteristics or random fluctuations. If a single measurement shows an extreme result but retesting shows regression toward the mean, the initial value likely reflected chance variation. Conversely, if repeated measurements consistently show treatment-induced changes across a group's average, this provides stronger evidence of genuine treatment effectiveness rather than statistical artifact.