ANOVA Calculator

Background.

Do three groups really differ, or did sampling luck spread their averages apart? Analysis of variance answers by comparing two kinds of spread: how far the group means sit from the overall mean (between-group variation) against how much observations scatter inside their own groups (within-group variation). Their ratio — each first divided by its degrees of freedom — is the F statistic this page computes, along with every intermediate quantity a hand check or a homework write-up needs.

The intuition is a signal-to-noise ratio. If the group means differ only because of ordinary sampling noise, the between-group spread should be about the same size as the within-group spread, and F lands near 1. Real differences between groups inflate the numerator only, pushing F upward — an F of 14 says the between-group signal is fourteen times the noise floor, far beyond what chance produces with these sample sizes.

Fisher's construction solved a real problem that repeated t-tests cannot: comparing three groups pairwise runs three tests and compounds their false-positive rates, while ANOVA asks one question — are all means equal? — with one test at one significance level. A significant F licenses the follow-up question of which groups differ (post-hoc comparisons); a nonsignificant one ends the inquiry early.

This page implements the classic one-way, three-group case: paste each group's observations and read F, the degrees of freedom pair, and both sums of squares. The surrounding statistical machinery — p-values from the F distribution, equal-variance assumptions and their Welch remedy, repeated measures, factorial designs — stays with your statistics package, as the scope note says.

What is anova calculator?

One-way ANOVA (analysis of variance) is the standard test of whether several group means are all equal. It decomposes total variation into a between-groups sum of squares (SS_between, driven by group means straying from the grand mean) and a within-groups sum of squares (SS_within, the scatter of observations around their own group means), then forms F = (SS_between/(k−1)) / (SS_within/(N−k)) for k groups and N total observations. Under the null hypothesis of equal means, F hovers near 1; large values are evidence the groups genuinely differ, judged against the F distribution with (k−1, N−k) degrees of freedom.

How to use this calculator.

1. 01Enter each group's observations as comma-separated numbers — the groups need not have equal sizes, though this page expects exactly three groups.
2. 02Read the F statistic and the two degrees of freedom: df_between = k−1 = 2 here, df_within = N−k.
3. 03Look up (or compute in any statistics tool) the p-value for F with those degrees of freedom — at df (2, 9), the 5% critical value is 4.26, so any F above it rejects equal means at that level.
4. 04Use SS_between and SS_within to report effect size: η² = SS_between/(SS_between+SS_within) is the share of total variation explained by group membership.
5. 05Treat a significant F as an invitation, not a conclusion: it says the means are not all equal — identifying which pairs differ needs post-hoc tests (Tukey HSD and kin), and the validity of all of it rests on independent observations with roughly equal group variances.

The formula.

The decomposition is exact bookkeeping: for every observation, (x − grand mean) = (x − group mean) + (group mean − grand mean), and squaring and summing makes the cross-terms vanish, so SS_total = SS_within + SS_between. Each sum of squares is then scaled by its degrees of freedom — k−1 for between (k group means, one constraint from the grand mean), N−k for within (N observations, k fitted group means) — turning sums into mean squares, which are variance estimates. Under the null hypothesis both mean squares estimate the same error variance, so their ratio F = MS_between/MS_within is distributed as F(k−1, N−k) and concentrates near 1; group-mean differences add a positive term to the numerator's expectation only, which is why the test is one-tailed upward. The engine parses the groups, forms the means and both sums of squares, and evaluates F in Decimal arithmetic, rounding once at the output — the textbook computation, with the intermediate quantities exposed rather than hidden.

A worked example.

Frequently asked questions.

References& sources.

1. [1]NIST/SEMATECH. e-Handbook of Statistical Methods, 2012. Chapter 1.3 and Chapter 4. Retrieved 2026-08-06. independence: primary; access: open.
2. [2]OpenStax, Rice University. Introductory Statistics 2e, 2023. Chapters 10–13. Retrieved 2026-08-06. independence: secondary-check; access: open.
3. [3]OpenStax, Rice University. College Algebra 2e, 2021. Algebra and function chapters. Retrieved 2026-08-06. independence: primary; access: open.
4. [4]OpenStax, Rice University. Principles of Data Science, 2024. Section 4.4, Analysis of Variance (ANOVA). Retrieved 2026-08-06. independence: secondary-check; access: open.
5. [5]National Institute of Standards and Technology. NIST SI Guide, 2008 edition. Appendix B unit relationships. Retrieved 2026-08-06. independence: primary; access: open.