Estimate the population variance and standard deviation with a confidence interval based on the chi-square distribution. This is useful for quality control and precision studies.
[ (n − 1)s² ÷ χ²(α/2, upper), (n − 1)s² ÷ χ²(α/2, lower) ]
The interval is not symmetric around s², because the chi-square distribution is skewed. It relies strongly on the data being normally distributed.
Example
A sample of 25 has s = 2.8. With 24 df, the 95% chi-square critical values are 39.36 and 12.40, so the interval for σ² is [24 × 7.84 ÷ 39.36, 24 × 7.84 ÷ 12.40] = [4.78, 15.17], and for σ it is [2.19, 3.90]. Notice that the interval is not centred on the estimate.