Confidence Interval for Variance Calculator

Your sample

95% confidence interval for the variance
[39.0203, 123.8595]
The interval is asymmetric: it stretches further above s² = 64.0000 than below it — the chi-square distribution is skewed, so the interval is not s² ± a margin.
CI for σ
[6.2466, 11.1292]
Lower critical χ²(0.025)
12.4012
Upper critical χ²(0.975)
39.3641
Degrees of freedom
24
Sample variance s²
64.0000
Sample size n
25

How the interval was calculated

    Chi-square curve

    The middle 95% of the χ²(24) curve lies between 12.4012 and 39.3641. The interval for σ² is built from those two cut-offs: divide (n − 1)s² by each of them.

    A confidence interval for a variance turns a sample’s spread into a range of plausible values for the population variance. Enter the sample size and standard deviation — or paste the raw data — and the calculator returns the interval for σ² and for σ, with the exact chi-square critical values and a shaded chi-square curve. For the example below, the 95% interval for σ² is [39.0203, 123.8595].

    How to use the confidence interval for variance calculator

    Use Summary stats when you already know the sample size and standard deviation. Use Raw data when you have the individual measurements — the calculator works out n and s for you.

    1. Choose a mode: Summary stats or Raw data.
    2. In Summary stats mode, enter the sample size n and the sample standard deviation s. In Raw data mode, paste the values, separated by commas, spaces or line breaks.
    3. Choose the confidence level: 90%, 95% or 99%.
    4. Read the headline interval for σ², then check the tiles for the interval for σ, the two chi-square critical values, the degrees of freedom and the sample variance. The calculation ledger shows every step, and the chart shades the middle of the chi-square curve.

    Try it now: with n = 25 and s = 8, the 95% interval for σ² is [39.0203, 123.8595] and for σ it is [6.2466, 11.1292]. Paste the raw list 12, 15, 14, 18, 16, 20, 22, 19, 17, 21 and it gives s = 3.2042 with a 95% interval for σ² of [4.8573, 34.2173].

    What is a confidence interval for a variance?

    When the data come from a normal population, the quantity (n − 1)s²/σ² follows a chi-square distribution with n − 1 degrees of freedom. That fact gives the interval:

    (n − 1)s² / χ²(1 − α/2) < σ² < (n − 1)s² / χ²(α/2)

    where χ²(p) is the p-quantile of the chi-square distribution with n − 1 degrees of freedom — the same distribution behind the chi-square test and the F-distribution. Note that the two divisions run the other way round from what you might expect: the lower end of the interval divides by the upper critical value, and the upper end divides by the lower critical value. Dividing by a larger number gives a smaller result.

    Why the interval is asymmetric

    A confidence interval for a mean is symmetric: estimate ± margin. An interval for a variance is not. The chi-square distribution is skewed to the right — it has a long tail on the high side — so the two critical values are not the same distance from the centre, and the interval stretches further above s² than below it. With n = 25 and s = 8, s² = 64: the interval [39.0203, 123.8595] reaches 24.9797 below 64 but 59.8595 above it. That asymmetry is a property of the variance, not a rounding artefact — and it is why you should never report a variance interval as s² ± something.

    The interval for σ is the square root of the interval for σ²

    If [L, U] is the interval for σ², then [√L, √U] is the interval for σ. Nothing else changes: the same critical values, the same confidence level. For the example, √39.0203 = 6.2466 and √123.8595 = 11.1292, so the 95% interval for σ is [6.2466, 11.1292]. Report the σ interval when your readers think in standard deviations, and the σ² interval when the variance itself is the quantity of interest.

    The interval for σ is just the square root of the interval for σ² — so always build the σ² interval first, then take square roots.

    Worked examples

    Example 1: summary statistics (n = 25, s = 8, 95%)

    The degrees of freedom are df = 25 − 1 = 24. The critical values are χ²(0.025) = 12.4012 and χ²(0.975) = 39.3641. The numerator is (n − 1)s² = 24 × 64 = 1536, so the interval is:

    Lower end: 1536 / 39.3641 = 39.0203. Upper end: 1536 / 12.4012 = 123.8595.

    So the 95% confidence interval for σ² is [39.0203, 123.8595], and for σ it is [6.2466, 11.1292]. At 90% the critical values are χ²(0.05) = 13.8484 and χ²(0.95) = 36.4150, giving [42.1804, 110.9151] for σ² — narrower, as a lower confidence level always gives.

    Example 2: raw data (10 measurements, 95%)

    The measurements are 12, 15, 14, 18, 16, 20, 22, 19, 17, 21. Their mean is 17.4, and the sum of squared deviations from the mean is 92.4, so the sample variance is s² = 92.4 / 9 = 10.2667 and s = 3.2042. With df = 9, the critical values are χ²(0.025) = 2.7004 and χ²(0.975) = 19.0228, and the numerator is (n − 1)s² = 92.4:

    Lower end: 92.4 / 19.0228 = 4.8573. Upper end: 92.4 / 2.7004 = 34.2173.

    So the 95% confidence interval for σ² is [4.8573, 34.2173] and for σ it is [2.2039, 5.8496]. Note how wide it is relative to s² = 10.2667 — with only 10 observations the variance is poorly pinned down.

    Edge cases

    Small samples. With n = 10 and s = 5, the 95% interval for σ² is [11.8279, 83.3213] — the upper end is more than seven times the lower end. Variance intervals are wide when n is small; that is honest, not a flaw.

    Two observations. With n = 2 the interval exists (df = 1) but it is extremely wide, and the calculator warns you rather than refusing — treat it as a rough guide, not a precise range.

    Identical values. If every value is the same, s = 0 and the interval collapses to [0, 0]. A confidence interval needs variation in the data.

    Non-normal data. The interval assumes the population is normal, and it is sensitive to that assumption — much more so than a confidence interval for a mean. If a Shapiro-Wilk test suggests the data are not normal, treat the interval with caution. Skew or heavy tails in the data distort the interval more than they distort a confidence interval for a mean.

    Frequently asked questions

    Why isn’t the interval symmetric?

    Because the chi-square distribution is skewed to the right, so the two critical values are not equally spaced around the centre, and the interval is built by dividing by them rather than by adding and subtracting a margin. The lower end divides by the upper critical value and the upper end divides by the lower one, so the interval stretches further above s² than below it. With n = 25 and s = 8, the interval [39.0203, 123.8595] reaches 24.9797 below s² = 64 but 59.8595 above it.

    Do I need normal data?

    Yes, in principle: the interval is derived from the fact that (n − 1)s²/σ² follows a chi-square distribution, which holds exactly only when the population is normal. The interval is sensitive to non-normality — heavy tails or skew in the data distort it more than they distort a confidence interval for a mean. If you are unsure, run a Shapiro-Wilk test first and treat the interval as approximate if the data look non-normal.

    The interval on this page is the standard chi-square interval for a population variance. All worked examples on this page were recomputed with Python (scipy.stats.chi2). Last updated 19 November 2026.

    See all 91 free statistics tools.