F-Distribution Calculator

F-Distribution Calculator

Find the probability or critical value for any F statistic, or run the F-test for two variances โ€” with the shaded F curve and the worked steps.

Numerator degrees of freedom, df₁ ≥ 1.
Denominator degrees of freedom, df₂ ≥ 1.
The F statistic, x ≥ 0.
The critical value F* shown below is the right-tail cut-off at this α.
F PROBABILITY ยท DF 3 / 12
0.9654

P(F ≤ 4.00) = 0.9654 โ€” about 96.5% of the F distribution with 3 and 12 degrees of freedom lies below 4.00.

P(F ≤ x)
0.9654
P(F ≥ x)
0.0346
Critical F (α = 0.05)
3.4903
Mean of F
1.2000
Variance of F
1.5600
Degrees of freedom
3 / 12

Calculation

  1. Distribution: F with df1 = 3 and df2 = 12.
  2. P(F ≤ 4.00) = I0.5000(3/2, 12/2) = 0.9654, where t = df1·x/(df1·x + df2) = 0.5000.
  3. P(F ≥ 4.00) = 1 − 0.9654 = 0.0346.
  4. Critical value: F* = 3.4903 at α = 0.05 (right tail, df 3/12).
  5. Mean of F = df2/(df2−2) = 12/10 = 1.2000.
  6. Variance of F = 2·df2²·(df1+df2−2)/(df1·(df2−2)²·(df2−4)) = 1.5600.

Shaded F curve โ€” left-tail probability

The shaded area to the left of 4.00 is the probability P(F ≤ 4.00) = 0.9654. The dot marks F = 4.00.

F* = 3.4903 at α = 0.05: an F statistic above 3.4903 falls in the top 5% of this distribution. Mean of F = 1.2000; variance of F = 1.5600.

The F-distribution calculator gives the probability and the critical value for any F statistic, and runs the F-test for two variances. Enter the degrees of freedom, read the p-value from the shaded curve, and check the worked steps. The F-test compares two sample variances; a large ratio with a small p-value means the variances differ.

How to use the F-distribution calculator

The calculator has three modes. Choose the one that matches your question, enter the numbers, and read the headline result — then check the shaded F curve and the worked steps underneath.

  1. Choose a mode: F Probability for an area under the curve, Critical Value for the cut-off at a chosen α, or F-Test (Two Variances) to compare two sample variances.
  2. Enter the degrees of freedom df₁ and df₂ — or, for the F-test, the two sample variances and sample sizes.
  3. Pick the tail you need (left tail, right tail, or between two values) and, for the critical value, the significance level α.
  4. Read the headline result, then check the results grid for the p-value, critical value, mean and variance. The calculation ledger shows every step, and the chart shades the region you asked about.

Try it now: with df₁ = 3, df₂ = 12 and x = 4, the calculator gives P(F ≤ 4) = 0.9654, P(F ≥ 4) = 0.0346 and a critical value of 3.4903 at α = 0.05.

What is the F-distribution used for?

The F-distribution is the distribution of the ratio of two independent chi-square variables, each divided by its degrees of freedom. If U follows a chi-square distribution with df₁ degrees of freedom and V follows one with df₂, independently, then F = (U/df₁)/(V/df₂) follows an F-distribution with df₁ and df₂ degrees of freedom. It is named after R. A. Fisher; G. W. Snedecor popularised it, which is why it is sometimes called the Fisher–Snedecor distribution.

The F-distribution lives on [0, ∞), is right-skewed, and its shape depends on both degrees of freedom. It is the workhorse distribution for comparing variances:

  • Comparing two variances — the F-test on this page: the ratio of two sample variances follows an F-distribution when the null hypothesis of equal variances is true.
  • One-way ANOVA — the test statistic for comparing three or more group means is an F ratio; use the one-way ANOVA calculator for that test.
  • Regression — the overall F-test checks whether a regression model explains significantly more variation than a model with no predictors.
  • Comparing nested models — the extra-sum-of-squares F-test compares a simpler model with a fuller one.

Because the F statistic is a ratio of two variances, it can never be negative, and large values count against the null hypothesis. The t-distribution calculator covers the analogous distribution for comparing means.

F-distribution mean and variance

The mean and variance of the F-distribution depend only on the denominator degrees of freedom, df₂ — and they do not always exist:

  • Mean = df₂/(df₂ − 2), when df₂ > 2. If df₂ ≤ 2, the mean does not exist.
  • Variance = 2·df₂²·(df₁ + df₂ − 2) / (df₁·(df₂ − 2)²·(df₂ − 4)), when df₂ > 4. If df₂ ≤ 4, the variance does not exist.

As df₂ grows, the mean approaches 1 and the variance shrinks towards 0 — the distribution concentrates around 1, which is why F ratios near 1 are unremarkable and large F ratios are not.

df₁df₂MeanVariance
3121.20001.5600
5201.11110.7099
2301.07141.3246
10101.25000.9375
4401.05260.6464

The F-test for two variances

The F-test compares the variances of two independent samples. The test statistic is the ratio of the larger sample variance to the smaller one, so F is always at least 1:

F = larger s² / smaller s², with df₁ = nlarge − 1 and df₂ = nsmall − 1. The p-value is two-sided by default: 2 × the smaller tail probability.

The test assumes:

  • Both samples are random and independent of each other.
  • Each population is normally distributed — the F-test is sensitive to non-normality, much more so than the t-test, so check the shape of your data first.
  • Observations are measured on a continuous scale.

Decision rule: reject the null hypothesis of equal variances when F falls outside the two-sided critical values at your α, or equivalently when the two-sided p-value is below α. The one-sided p-value is the right-tail probability P(F ≥ observed F) — use it only when you specified a direction before seeing the data. The F-test on this page is two-sided by default, which is the safe choice when you are testing for any difference rather than a difference in a pre-specified direction.

The F-test is the companion of the one-way ANOVA: ANOVA generalises the variance-ratio idea to three or more groups, and its test statistic is also an F ratio. For categorical data, use the chi-square test calculator instead.

Worked examples

Example 1 — F probability with df₁ = 3 and df₂ = 12

Find P(F ≤ 4), P(F ≥ 4) and the 5% critical value for an F-distribution with 3 and 12 degrees of freedom. This is the worked example used by Restatica. The calculator gives:

  • P(F ≤ 4) = 0.9654
  • P(F ≥ 4) = 1 − 0.9654 = 0.0346
  • F* = 3.4903 at α = 0.05
  • Mean = 12/10 = 1.2000 and variance = 1.5600

Because the right-tail probability 0.0346 is below 0.05, an observed F of 4 would be significant at the 5% level.

Example 2 — F-test for two variances

Two samples have variances s₁² = 25 (n₁ = 10) and s₂² = 10 (n₂ = 12). Test whether the population variances differ at α = 0.05.

  • F = 25/10 = 2.5000, with df₁ = 10 − 1 = 9 and df₂ = 12 − 1 = 11.
  • One-sided p = P(F ≥ 2.5000) = 0.0771; two-sided p = 2 × 0.0771 = 0.1542.
  • Critical values at α = 0.05 (two-sided): [0.2556, 3.5879].
  • Decision: F = 2.5000 lies between the critical values and p = 0.1542 > 0.05, so fail to reject H₀ — there is no evidence the variances differ.

Frequently asked questions

What is the F-distribution used for?

The F-distribution describes the ratio of two independent chi-square variables, each divided by its degrees of freedom. It is used to compare variances (the F-test), to test three or more means (one-way ANOVA) and to test regression models. An F statistic is always non-negative, and large values count against the null hypothesis.

When does the F-distribution have a mean and a variance?

The mean exists only when df₂ > 2 and equals df₂/(df₂ − 2); the variance exists only when df₂ > 4. When df₂ is too small, the tails are so heavy that the mean or variance does not exist — the calculator says so instead of showing a number.

What does the F-test for two variances assume?

It assumes two independent random samples from normally distributed populations. The F-test is sensitive to non-normality: if the data are skewed or heavy-tailed, the p-value can be misleading even for large samples. Check the shape of the data before trusting the result.

Why is the F-test two-sided by default?

Because the question is usually whether the variances differ in either direction. The statistic is formed as the larger variance over the smaller one, so it is always at least 1; a two-sided test therefore needs both tails of the F-distribution (or twice the smaller tail). A one-sided test is only appropriate when you specified a direction before seeing the data.

Source: the critical F values quoted on this page are from the NIST/SEMATECH e-Handbook of Statistical Methods, Upper Critical Values of the F Distribution. All examples on this page were checked against that table.

See all 67 free statistics tools on Empirical Rule Calculators.