One-Way ANOVA & F Statistic Calculator

Compare the means of two to eight independent groups using classical one-way ANOVA. Work from individual observations or from each group's sample size, mean and sample standard deviation. Connect the F statistic and upper-tail p-value to the between- and within-group sums of squares, and see which groups contribute most to each component.

Comparing group means

One-way ANOVA explorer

See how differences between group means compare with variation inside the groups.

Enter independent groups

Enter each sample’s count, mean and sample standard deviation. Drafts are kept when you switch formats.

Group 1
Group 2
Group 3

Test and effect estimates

F statistic

12

Degrees of freedom: 2, 6

Upper-tail p-value

0.008

Under the equal-means model

Eta squared (η²)

0.8

Share of variation between groups

Omega squared (ω²)

0.70967742

Adjusted effect estimate

F(2, 6) = 12. Upper-tail p-value 0.008.

Share of total variation

Each bar starts at zero and shows that row’s share of total sum of squares.

Between groups Within groups
  • All groups

    Between 80% · Within 20%

  • Group 1

    Between 40% · Within 6.6666667%

  • Group 2

    Between 0% · Within 6.6666667%

  • Group 3

    Between 40% · Within 6.6666667%

This test assumes independent observations, approximately normal errors and a common population variance. The p-value addresses the equal-means hypothesis for all groups together.

ANOVA table

9 observations · Grand mean 4

SourceSum of squaresDegrees of freedomMean square
Between groups24212
Within groups661
Total308—

Group statistics and contributions

GroupCountMeanSample SDBetween SSWithin SS
Group 1321122
Group 234102
Group 3361122

How to use this calculator

  1. Choose Summary statistics or Raw observations. Start with the summary example or load the NIST resistor data to see a worked example.
  2. Name each independent group. With Summary statistics, enter the count, mean and sample SD for each group. With Raw observations, enter decimal values separated by commas or whitespace.
  3. Use Add group or Remove to work with two to eight groups. Review the F statistic, its two degrees of freedom, the upper-tail p-value and the two effect estimates.
  4. Compare the between- and within-group contributions in Share of total variation. Check the ANOVA table for the sums of squares, degrees of freedom and mean squares.
  5. Download ANOVA CSV or open the selectable export for the input values, calculation ledger and underlying chart shares.

The Formula

The mathematical logic powering this calculation is:

For k groups with total count N, the grand mean x̄ is Σnⱼx̄ⱼ/N. The between-group sum of squares is SSᵦ = Σnⱼ(x̄ⱼ − x̄)². The within-group sum is SS𝓌 = Σ(nⱼ − 1)sⱼ², where sⱼ is the sample standard deviation. The total sum is SSₜ = SSᵦ + SS𝓌.

The between- and within-group degrees of freedom are k − 1 and N − k. Dividing each component by its degrees of freedom gives its mean square. F = MSᵦ/MS𝓌 when MS𝓌 is positive. Under the equal-means null model, the p-value is the upper-tail probability of F with those two degrees of freedom. The calculation evaluates the tail using the regularized incomplete beta function.

η² = SSᵦ/SSₜ. The adjusted estimator is ω² = [SSᵦ − (k − 1)MS𝓌]/(SSₜ + MS𝓌). A negative ω² is retained as calculated. When SSₜ = 0, both effect estimates are undefined. When MS𝓌 = 0 and the means differ, the decomposition and effect estimates are available and the F statistic and p-value are unavailable.

Entered decimals are carried exactly through sums of squares, mean squares and effect ratios. Sample standard deviations from raw observations and F probabilities are numerical approximations. The CSV includes the exact fractions and the log₁₀ of the reported p-value for reproducibility.

Worked Example

If you input:

  • Input format: Summary statistics
  • Group 1 (count, mean, sample SD): 3, 2, 1
  • Group 2 (count, mean, sample SD): 3, 4, 1
  • Group 3 (count, mean, sample SD): 3, 6, 1

Result:

The grand mean is 4. The between-group and within-group sums of squares are 24 and 6. The degrees of freedom are 2 and 6, giving mean squares of 12 and 1. The F statistic is 12 and its upper-tail p-value is 0.008. Eta squared is 0.8 and omega squared is 22/31, about 0.7097.

Frequently Asked Questions

What does the ANOVA F statistic measure?

It compares the mean square between the group means with the mean square of observations around their own group means. Larger values correspond to more between-group variation relative to within-group variation. The numerator degrees of freedom are the group count minus one. The denominator degrees of freedom are the observation count minus the group count.

How should I interpret the p-value?

Under the equal-population-means hypothesis and its assumptions, the p-value is the probability of an F statistic at least as large as the one observed. It summarizes the comparison across all groups together. Identifying particular differences calls for a planned contrast or a follow-up procedure that accounts for multiple comparisons.

When is this form of one-way ANOVA appropriate?

Use a quantitative outcome grouped by one factor, with independent observations, an approximately normal error distribution within each group and a common population variance. A repeated-measures or matched design needs a method that accounts for dependence. Welch's ANOVA is a separate procedure for unequal variances.

Which values can I enter?

Enter 2 to 8 named groups. Raw mode takes 2 to 200 plain decimal observations per group, separated by commas or whitespace. Summary mode takes integer counts of 2 to 10,000, a mean and a nonnegative sample SD. Decimal magnitudes can be up to one billion, with up to six fractional places.

Should I use sample or population standard deviation?

Use each group's sample standard deviation, computed with denominator n − 1. If you have the original observations, raw mode computes it directly. Rounded summaries produce results based on their entered precision.

Why might F or p be unavailable?

If the observations within every group are identical, the within-group mean square is zero and the F ratio is undefined. The ANOVA table still reports the zero denominator. If the group means are identical as well, total variation is zero and variation shares and effect estimates are undefined.

Why can omega squared be negative?

Omega squared subtracts an adjustment for within-group variation. Sampling can make the adjusted numerator negative when the observed group means are close together. The calculator preserves that computed estimate. Eta squared is the observed proportion of total sum of squares attributed to differences between group means.

Similar Calculators