What is a Variance Confidence Interval Calculator?
A confidence interval for a variance gives a range of plausible values for a population variance σ² (and its square root, the standard deviation σ) from one sample. It divides (n − 1)s² by two chi-square critical values with n − 1 degrees of freedom: ((n − 1)s² / χ²_upper, (n − 1)s² / χ²_lower). The method assumes the population is normal, and the interval is not symmetric about s². For example, n = 25 and s = 2.4 give a 95% interval of (3.51, 11.15) for σ² and (1.87, 3.34) for σ.
What the Variance Confidence Interval Calculator does
Enter raw data, or the sample size with either the sample standard deviation or the sample variance, and choose a confidence level. The calculator returns the interval for σ² and for σ, the two chi-square critical values it used, a shaded chi-square curve, and the same interval at 90%, 95% and 99%. Choose Upper bound only or Lower bound only when the question is one-sided, such as showing that a process standard deviation is no larger than a specification.
It goes by several names: a confidence interval for variance calculator, a confidence interval for standard deviation calculator (or standard deviation confidence interval calculator), and a chi-square confidence interval calculator, after the distribution the limits come from. They all do the same job: one sample in, a range for σ² and σ out, at 95% or any other confidence level.
Use it when the quantity of interest is variability itself: the consistency of fill weights, the precision of a measuring instrument, the volatility of returns, or the spread of test scores. An interval is usually more informative than a bare s, because standard deviations estimated from small samples are much less certain than most people expect. To test a specific claimed value of σ instead, use the One-Sample Variance Test Calculator; to compare the variances of two groups, use the F-Test Calculator.
How to use the Variance Confidence Interval Calculator
- Set the Confidence level (95% unless your field uses another).
- Choose the Interval type: two-sided for a range, or an upper or lower bound for a one-sided statement.
- Choose Summary statistics and enter Sample size, n with either the standard deviation s or the variance s², or choose Raw data and paste the observations.
- Read the interval for σ² and for σ. The standard deviation interval is the one to quote when you want the original units.
- Check the normality warning: this method is sensitive to non-normal data at every sample size.
How to calculate a confidence interval for a variance
Five steps take you from a sample to the interval. The calculator runs the same steps and prints them in the step-by-step solution, so you can check a hand calculation against it.
- Find the sample variance: s² with the n − 1 divisor. If you have the standard deviation, square it: s = 2.4 gives s² = 5.76.
- Find the degrees of freedom and (n − 1)s²: df = n − 1. For n = 25, df = 24 and (n − 1)s² = 24 × 5.76 = 138.24.
- Look up two chi-square critical values with n − 1 df, leaving α/2 in each tail. At 95% confidence with 24 df they are 12.4012 and 39.3641.
- Divide (n − 1)s² by each critical value. The larger value gives the lower limit: 138.24 / 39.3641 = 3.5118, and 138.24 / 12.4012 = 11.1474.
- Take square roots for the standard deviation: √3.5118 = 1.8740 and √11.1474 = 3.3388.
Confidence interval for variance formula
If the data come from a normal population, (n − 1)s² / σ² follows a chi-square distribution with n − 1 degrees of freedom. Solving the central probability statement for σ² gives the interval.
| Symbol | Meaning |
|---|---|
| s² | sample variance (n − 1 divisor) |
| n | sample size; degrees of freedom df = n − 1 |
| α | 1 − confidence level |
| χ²₍p₎ | value with cumulative probability p under the chi-square distribution with n − 1 df |
Subscripts here are left-tail (cumulative) probabilities. Many textbooks and printed tables index χ² by the area to the right, so the same two numbers appear there as χ²₍α/2₎ (the large one) and χ²₍1−α/2₎ (the small one). Whichever notation you use, the larger critical value produces the lower limit.
Look up the critical values in the chi-square table or with the Chi-Square Critical Value Calculator.
This is the equal-tailed interval used by nearly all textbooks and software. It is not the shortest possible interval, because the chi-square distribution is skewed.
The same formula appears in the NIST/SEMATECH e-Handbook of Statistical Methods (section 1.3.5.8), which also notes that the interval for σ is the square root of the interval for σ². The sensitivity of chi-square methods for variances to non-normal data was shown by G. E. P. Box (1953, Biometrika).
Worked example
Worked example: consistency of a filling machine
Problem. A quality engineer weighs 25 bottles from a filling line and finds a sample standard deviation of 2.4 g. Find a 95% confidence interval for the population variance and standard deviation of fill weight.
- 01Sample variance: s² = 2.4² = 5.76 g², and df = 25 − 1 = 24, so (n − 1)s² = 24 × 5.76 = 138.24.
- 02Critical values with 24 df: χ²(0.025) = 12.4012 and χ²(0.975) = 39.3641.
- 03Lower limit for σ²: 138.24 / 39.3641 = 3.5118. Upper limit: 138.24 / 12.4012 = 11.1474 (using the unrounded critical value).
- 04Square roots: √3.5118 = 1.8740 and √11.1474 = 3.3388.
Answer. With 95% confidence the variance of fill weight is between 3.51 and 11.15 g², and the standard deviation is between 1.87 and 3.34 g. Note the asymmetry: the interval runs 0.94 g above s = 2.4 but only 0.53 g below it. At 90% confidence the interval for σ narrows to (1.95, 3.16), and at 99% it widens to (1.74, 3.74).
90%, 95% and 99% confidence intervals for a variance
A higher confidence level moves the two chi-square critical values further apart, so both limits spread out and the upper limit moves faster than the lower one. The table uses the filling-machine example, n = 25 and s = 2.4 (df = 24). For other levels or sample sizes, use the Chi-Square Critical Value Calculator or the chi-square table.
| Confidence level | Lower χ² | Upper χ² | Interval for σ² | Interval for σ |
|---|---|---|---|---|
| 90% | 13.8484 | 36.4150 | (3.7962, 9.9824) | (1.9484, 3.1595) |
| 95% | 12.4012 | 39.3641 | (3.5118, 11.1474) | (1.8740, 3.3388) |
| 99% | 9.8862 | 45.5585 | (3.0343, 13.9831) | (1.7419, 3.7394) |
Confidence interval for standard deviation
There is no separate formula for a confidence interval for a standard deviation. Build the interval for σ² and take the square root of both limits; because the square root keeps the order of positive numbers, the confidence level carries over unchanged. In the filling-machine example the 95% interval for σ² is (3.5118, 11.1474) g², so the interval for σ is (√3.5118, √11.1474) = (1.87, 3.34) g.
Do not build it as s ± a margin of error: that ignores the skew and can even give a negative lower limit for a small sample. If only the worst case matters, as in a specification for process spread, use a one-sided upper bound instead. For n = 25 and s = 2.4 the 95% upper bound is σ ≤ 3.16 g, the same number as the upper limit of the two-sided 90% interval, because both put 5% in the left tail. Quote the interval for σ when you want an answer in the original units, and the interval for σ² when the variance itself feeds a later calculation such as an F-test or a pooled estimate.
Why the interval is not symmetric around s²
A confidence interval for a mean is the estimate plus or minus a margin. A variance interval cannot be built that way, because a variance cannot be negative and its sampling distribution is skewed to the right: a sample is more likely to miss the rare large deviations and understate σ² than to overstate it. The chi-square method handles this by dividing by two different critical values instead of adding and subtracting one margin.
The practical effect is that the upper limit is always further from s than the lower limit, and the gap is dramatic for small samples. The table shows the 95% interval for σ as a multiple of s. With n = 5 the true standard deviation could be almost three times the sample value.
| Sample size, n | df | Lower limit for σ | Upper limit for σ |
|---|---|---|---|
| 5 | 4 | 0.599 × s | 2.874 × s |
| 10 | 9 | 0.688 × s | 1.826 × s |
| 25 | 24 | 0.781 × s | 1.391 × s |
| 50 | 49 | 0.835 × s | 1.246 × s |
| 100 | 99 | 0.878 × s | 1.162 × s |
| 500 | 499 | 0.942 × s | 1.066 × s |
How much the normality assumption matters
The t interval for a mean is protected by the central limit theorem: with a large sample it works for almost any population. The chi-square interval for a variance has no such protection. Its width depends on the fourth moment (kurtosis) of the population, and it assumes the value that a normal distribution has.
For heavy-tailed populations the true sampling variability of s² is larger than the chi-square formula assumes, so a nominal 95% interval can cover the true variance far less often than 95% of the time, and collecting more data does not fix it. If your data show clear skew, outliers or heavy tails, treat the interval as approximate at best and consider a bootstrap interval for the standard deviation instead.
Common use cases
- Process capability and quality control: put a confidence bound on the standard deviation of a dimension, weight or fill volume.
- Measurement system analysis: state how precisely a gauge or assay’s repeatability has been estimated.
- Finance: attach an interval to an estimated volatility instead of quoting a single number.
- Study planning: use the upper bound for σ from a pilot study as a cautious input to a sample size calculation.
- Reporting: give σ with its interval in a methods section when variability is a result in its own right.
Assumptions and limitations
- Normal population required. The method is not robust to non-normality at any sample size. Check the shape of the data before relying on it.
- Random, independent observations. Trends, batches or repeated measurements on the same unit make s² understate or overstate the real variability.
- The sample variance must use the n − 1 divisor. If your software reports the population (÷ n) version, multiply it by n / (n − 1) first.
- Intervals for variances are wide. Even with n = 100 the 95% limits for σ run from 12% below s to 16% above it.
- For the ratio of two variances use the F-Test Calculator; for an interval on the mean of the same sample use the Mean Confidence Interval Calculator.
Frequently asked questions
How do you calculate a confidence interval for a variance?
What is the 95% confidence interval for a variance?
How do you find a confidence interval for a standard deviation?
Why is the chi-square distribution used for a variance interval?
Why is the confidence interval for a variance not symmetric?
Which chi-square value gives the lower limit?
Does a large sample make the variance interval valid for non-normal data?
When should I use a one-sided bound instead of a two-sided interval?
Cite this page
Statculator. (2026, November 14). Variance Confidence Interval Calculator. Retrieved November 14, 2026, from https://statculator.com/confidence-intervals/variance-confidence-interval-calculator/