Regression to the Mean

Regression to the mean is the statistical tendency for an unusually high or low measurement to be followed by one closer to the average.
555
minutes read

What Is Regression to the Mean?

Regression to the mean is the statistical tendency for an unusually high or low measurement to be followed by one closer to the average. It occurs whenever two measurements of the same thing are imperfectly correlated, which describes almost every variable researchers study.

The phenomenon is not caused by any intervention. It is a property of repeated measurement, and it is easily mistaken for a treatment effect, an improvement, or a decline.

‍

Why Extreme Results Drift Back Toward Average

Any observed score combines a stable underlying value with some degree of chance. That chance component includes measurement error from instruments or observers, plus random variation in the thing being measured, such as daily swings in blood pressure or mood.

When a score is extreme, chance has probably pushed it in that direction. On the next measurement, the chance component is unlikely to be as extreme again, so the score tends to move back toward the person's true level. The weaker the correlation between the two measurements, the stronger the drift.

‍

Where Regression to the Mean Shows Up in Research

The effect appears wherever participants are selected because of extreme values. Clinical trials that enroll patients with high cholesterol or severe symptoms, education programs that target the lowest-scoring students, and quality projects aimed at the worst-performing hospitals are all vulnerable.

It also matters when evidence is synthesized. Uncontrolled before-and-after studies with extreme-score entry criteria deserve extra caution during evidence screening for a systematic review, since their apparent effects may be partly statistical.

‍

Mistaking Regression to the Mean for a Real Effect

The classic error is attributing natural improvement to an intervention. Patients often seek care when symptoms peak, so many would feel better afterward with or without treatment. Part of what looks like a placebo effect in a trial can actually be this statistical drift.

The same mistake happens in policy. Speed cameras installed at sites after a spike in crashes may appear to cut accidents, even if some of the drop would have happened anyway. Without a comparison, the two explanations cannot be separated.

‍

Controlling for Regression to the Mean

Researchers can reduce or measure its influence by building safeguards into the study:

  1. Randomize participants to an intervention arm and a control group, so both regress equally and the difference between them reflects the intervention.
  2. Take more than one baseline measurement and use the average, which softens the impact of a single extreme reading.
  3. Select participants on one measurement, then use a separate, fresh measurement as the starting point for analysis.
  4. Adjust for baseline values statistically, for example with analysis of covariance (ANCOVA).
  5. Estimate the expected amount of regression from the test-retest correlation and compare it with the observed change.

‍

FAQs

Is regression to the mean a type of bias?

It is a statistical phenomenon rather than a bias in itself. However, it produces biased conclusions when researchers ignore it, especially in uncontrolled before-and-after studies. In that situation, natural movement toward the average gets misread as an effect of treatment, which overstates benefits and can lead to poor decisions.

Does regression to the mean happen in every study?

It can occur whenever measurements are imperfectly correlated and participants are chosen for extreme values. Its size depends on how extreme the selection was and how reliable the measurements are. In randomized studies it still occurs, but it affects all groups equally, so comparisons between groups remain valid.

Who first described regression to the mean?

Sir Francis Galton described the phenomenon in the 1880s. Studying the heights of parents and their adult children, he found that children of very tall or very short parents tended to be closer to average height. He called it “regression towards mediocrity,” and the idea later became central to modern statistics.

Can regression to the mean affect individual patients?

Yes. A patient measured on a bad day may show an unusually high reading that falls at the next visit without any real change in their condition. Clinicians reduce this risk by repeating measurements before diagnosing or changing treatment, and by treating single extreme readings with caution.

Does mean reversion mean the same thing?

The ideas are closely related. Mean reversion is the term finance and economics typically use for prices or returns that tend to move back toward a long-term average over time. In research methods, regression to the mean usually refers to the statistical effect of repeated measurement on selected extreme values.

Free access for academic researchers

Create your free QED account to validate your research, strengthen grant proposals, and uncover scientific insights.
We've sent you an access link.
Please check your inbox.

Didn't get your email? Check your spam folder or reach out to info@qedscience.com

Oops! Something went wrong while submitting the form.
Looking for QED for pharma, biotech or life science organizations?