Confidence interval
Also known as: CI · 95% CI · 95% confidence interval · Margin of error
A range around an estimated number that shows the uncertainty of the estimate. A 95% CI is built so that, if the experiment were repeated many times, 95% of the ranges would contain the true value.
Legal basis
NIST/SEMATECH e-Handbook — Confidence limits
Every metric measured on a sample is a guess with a margin of error — the observed Gini is an estimate, not the population's true Gini. The confidence interval is what turns a bare number into an honest statement: instead of a Gini of 32%, a Gini of 32% that lies between 27% and 37% with 95% confidence.
That is why, in a validation test, the central value decides and the CI is always reported. A Gini of 32% with a band of 30% to 34% is one thing; the same 32% with a band of 10% to 50% is almost knowing nothing. When a metric has no closed-form formula for its margin, the interval usually comes from bootstrap.
Frequently asked questions
What does a 95% CI mean?
It means the construction method captures the true value 95% of the time if the experiment were repeated many times. It is a measure of the estimate's uncertainty, not the probability that one specific number is correct.