Confidence intervals
What a 95% confidence interval really means and how to use intervals for means, proportions and Cpk.
A confidence interval gives a range of plausible values for a population parameter, based on a sample. A 95% interval is built by a method that captures the true value in 95% of samples.
Interval for a mean
Example: 25 measurements, mean 50.12, s = 0.30. With t ≈ 2.064 for 24 degrees of freedom, the 95% interval is , or 49.996 to 50.244.
Interval for a proportion
For large samples, . For few defectives or small samples use an exact (Clopper-Pearson) or Wilson interval.
Interval for a standard deviation
Based on the chi-square distribution. Standard deviations are much less precise than means: with 30 values, the 95% interval for runs roughly from 0.80s to 1.34s.
Intervals vs tolerance intervals vs prediction intervals
- Confidence interval: where the true mean (or other parameter) is.
- Prediction interval: where the next single value will fall.
- Tolerance interval: a range that contains a stated share of all values (for example 99% of parts) with stated confidence. This is the one to compare with specification limits.
Questions and answers
Does a 95% interval mean there is a 95% chance the true mean is inside?
Strictly, no. The true mean is fixed; the interval varies from sample to sample. In practice, treating it as "we are 95% confident" is fine.
Why should I care about the confidence interval on Cpk?
With 30 parts, a Cpk of 1.40 might really be anywhere from about 1.0 to 1.8. A customer requiring 1.33 may reject a study whose lower bound falls below 1.33. More data narrows the interval.
