# A confidence interval for the mean is not your personal range

Draft • Research checked 4 October 2026 • AI-assisted DomDNA editorial content. No independent clinical review or publication is claimed. General education, not individual medical advice.

A study reports an average improvement with a narrow confidence interval. It is tempting to read the two endpoints as a range for what you might experience. That interpretation usually gives the interval a job it does not have.

The interval concerns uncertainty about a group-level quantity under a statistical model. Individual experiences can vary much more widely, and the study's group may differ from the people applying the result.

## Find the quantity before reading the brackets

An interval can accompany a mean, a difference between means, a risk ratio or another estimate. Its meaning depends on that quantity. The brackets alone do not tell you which question was analysed.

[NIST's explanation of confidence limits for a mean](https://itl.nist.gov/div898/handbook/eda/section3/eda352.htm) describes uncertainty associated with estimating a population mean from a sample. In the conventional frequentist framework, the coverage refers to the behaviour of the method over repeated samples, under its assumptions.

That is not the same as saying that a particular person has a specified chance of experiencing a result inside the displayed interval. Nor is it a range containing most participants' actual measurements.

## A fictional travel-time study

Imagine researchers evaluate two appointment systems and estimate that one reduces average waiting time by four minutes. The hypothetical confidence interval for the mean difference runs from two to six minutes.

Some participants might wait longer, some might wait much less and some might see little change. The two-to-six-minute interval is not a promise that each person's waiting time improves by that amount. It describes uncertainty in the estimated average difference.

Now imagine a larger study produces a narrower interval around the same average. That could make the group estimate more precise without making individual waiting times more similar. Precision of an average and consistency of individual experiences are separate properties.

These figures are invented arithmetic illustrations. They are not estimates for any clinic, intervention or personal outcome.

## Other intervals answer other questions

A prediction interval and a tolerance interval can concern different targets from a confidence interval for a mean. [NIST's tolerance-interval guidance](https://www.itl.nist.gov/div898/handbook/prc/section2/prc263.htm), for example, discusses coverage of a population distribution.

Seeing a wider interval therefore does not automatically mean that a report used an inferior analysis. It may be answering a different question. Before comparing interval widths, check that the quantities and interval types are comparable.

Even an interval aimed at individual variation does not necessarily become a reliable personal forecast. The population, measurements, model and future setting still matter. A statistical label cannot remove uncertainty about how well a study applies elsewhere.

## Narrow does not mean free of bias

A precise estimate can still describe a biased comparison or a poorly measured outcome. Increasing sample size cannot by itself fix selective enrolment, a changed endpoint or an unsuitable measurement.

The [Greenland and colleagues methodological guide](https://pubmed.ncbi.nlm.nih.gov/27209009/?dopt=Abstract) emphasises assumptions and the hazards of statistical shortcuts. Its indexed abstract supports that general caution; this article does not attribute specific numbered interpretations from an inaccessible full-text section.

When a summary calls a result “certain” because its interval is narrow, separate sampling uncertainty from the other uncertainties. The calculation usually does not include every possible design problem or every difference between the research population and you.

## A useful note to keep beside a study

Copy the name of the estimated quantity, its units and the interval type. Then write what the interval does not describe. For the fictional waiting-time study, that note would read: “Uncertainty in the average difference, not the range of individual improvements.”

Next, look for how the report describes variation in actual participant outcomes. That information may be elsewhere, such as a distribution, a standard deviation or a plot. It answers a complementary question rather than competing with the confidence interval.

If individual variation is not reported in the material you have, leave it unknown. Do not manufacture it from the interval around the mean.

This small distinction makes research more useful: you can recognise a precise group estimate while resisting a false personal promise. It also helps you ask a focused follow-up question about variability instead of assuming the brackets already contain the answer.

For an educational starting point, the [DomDNA lifestyle quiz](https://domdna.com/quiz) records lifestyle answers, not DNA analysis or a clinical diagnosis.