One-Way ANOVA Calculator
Run the full F-test for three or more groups — from raw data or summary statistics — with the ANOVA table, eta squared, omega squared, Tukey HSD and Levene’s test.
Paste each group’s values, separated by commas, spaces or line breaks. Each group needs at least 2 values. Groups beyond the count you chose are ignored.
| Group | Mean | SD | n |
|---|---|---|---|
| A | |||
| B | |||
| C | |||
| D | |||
| E | |||
| F | |||
| G | |||
| H |
Tukey HSD and Levene’s test need the raw values, so they are not available from summary statistics. Enter a mean, SD and n for each group.
Enter your groups and press Calculate ANOVA.
ANOVA table
Tukey HSD — pairwise comparisons
Calculation
Shaded F curve — rejection region
The curve is the F-distribution for the ANOVA degrees of freedom. The shaded area is the rejection region beyond the critical value F*.
The one-way ANOVA calculator runs the full F-test for three or more groups from raw data or summary statistics. It prints the complete ANOVA table, eta squared and omega squared, Tukey HSD for the pairwise comparisons, and Levene’s test for equal variances. Enter your groups and read the verdict.
How to use the one-way ANOVA calculator
Use Raw data when you have the individual measurements for each group — paste each group’s values, separated by commas, spaces or line breaks. Use Summary stats when you already know each group’s mean, standard deviation and size.
- Choose a mode: Raw data or Summary stats.
- Set the number of groups (3 to 8) and the significance level (α).
- Enter each group’s values — or its mean, SD and n in Summary stats mode. Each group needs at least 2 values.
- Read the headline F statistic and p-value, then check the ANOVA table, eta squared, omega squared, Cohen’s f, the Tukey HSD comparisons and Levene’s test. The calculation ledger shows every step, and the chart shades the rejection region.
Try it now: with groups A = 6, 8, 7, 9, 8; B = 5, 4, 6, 5, 7; and C = 9, 10, 8, 9, 12, the calculator gives F = 13.5102 with df (2, 12) and p = 0.0008 — significant at the 5% level.
On a TI-84: press STAT, choose EDIT, and enter each group in a list (L1, L2, L3, …). Then press STAT, arrow to TESTS, choose H:ANOVA(, and enter the lists — ANOVA(L1,L2,L3). The calculator prints F and p.
When should you use a one-way ANOVA?
Use a one-way ANOVA when you have three or more groups and want to know whether their means differ. It is the right test when:
- There is one categorical factor (the grouping variable) and one continuous outcome.
- The groups are independent — each observation belongs to exactly one group.
- The data in each group are approximately normal, or each group is large enough that the central limit theorem applies.
- The variances are roughly equal across groups — Levene’s test on this page checks that for you.
If you have only two groups, a two-sample t-test is the simpler choice — and for two groups the two tests agree exactly (F = t²). If you are not sure which test fits your data, see which statistical test should I use?. To compare ANOVA with a t-test in detail, see ANOVA vs t-test.
This calculator runs a one-way ANOVA only. It does not run a two-way ANOVA (two factors at once).
Reading the ANOVA table
The ANOVA table splits the total variation into a between-groups part and a within-groups part:
| Source | SS | df | MS | F | p |
|---|---|---|---|---|---|
| Between groups | SSB | k − 1 | MSB | MSB ÷ MSW | from F(dfB, dfW) |
| Within groups | SSW | N − k | MSW | — | — |
| Total | SST | N − 1 | — | — | — |
- SSB measures how far the group means sit from the grand mean. SSW measures the scatter inside the groups. SST = SSB + SSW.
- df between = k − 1 (k groups); df within = N − k (N total observations).
- MSB = SSB ÷ df between; MSW = SSW ÷ df within. F = MSB ÷ MSW.
- A large F with a small p-value means the group means differ more than chance alone would explain.
The p-value comes from the F-distribution with (df between, df within) degrees of freedom. You can look up any F critical value with the F-distribution calculator.
For two groups only, ANOVA and the two-sample t-test agree exactly: F = t², and the p-values match.
Eta squared, omega squared and Cohen’s f
A significant p-value says the means differ; these three numbers say by how much.
- Eta squared (η²) = SSB ÷ SST — the share of the total variation that sits between groups. In Example 1 below, η² = 0.6925, so about 69% of the variation is between groups. Roughly, 0.01 is small, 0.06 is medium and 0.14 is large.
- Omega squared (ω²) = (SSB − dfB × MSW) ÷ (SST + MSW) — the same idea, corrected for the bias in η². It is the better estimate of the population effect, and it is always a little smaller than η².
- Cohen’s f = √(η² ÷ (1 − η²)) — the effect size on the standard-deviation scale. Roughly, 0.10 is small, 0.25 is medium and 0.40 is large. In Example 1, f = 1.5006, a very large effect.
Report η² or ω² alongside the p-value: a tiny p-value with a tiny effect size can mean a real but trivial difference.
Tukey HSD — what happens after a significant ANOVA
A significant ANOVA says at least one mean differs, but not which ones. Tukey’s HSD (honestly significant difference) compares every pair of groups while holding the overall error rate at α.
- For each pair it computes HSD = q × √(MSW ÷ n), where q is the studentised range critical value for k groups and df within. A pair is significant when the difference between its means is larger than the HSD.
- With unequal group sizes the calculator uses the Tukey–Kramer form: HSD = q × √(MSW ÷ 2 × (1 ÷ ni + 1 ÷ nj)).
- Tukey HSD needs the raw values, so it appears in Raw data mode only. In Summary stats mode the page says so and shows the ANOVA table without the pairwise comparisons.
In Example 1 the critical q is 3.7729 and the HSD is 2.1564, so A−B (difference 2.2000) and B−C (difference −4.2000) are significant, while A−C (difference −2.0000) is not.
Levene’s test and equal variances
ANOVA assumes the groups have equal variances. Levene’s test checks that assumption — on this page it uses the Brown–Forsythe form, which measures each value’s absolute deviation from its group median (medians, not means, so a stray outlier does not dominate the test).
- If Levene’s test does not reject (p > α), the equal-variance assumption is plausible and the ANOVA result stands. In Example 1, Levene’s F = 0.0690 with p = 0.9337, so the variances are plausibly equal.
- If Levene’s test rejects (p ≤ α), the variances differ and the ordinary ANOVA p-value is unreliable. The usual remedies are Welch’s ANOVA (which does not assume equal variances) or the Kruskal–Wallis test (a rank-based alternative). Neither is on this page.
- A group with zero variance (all its values identical) does not stop the calculator — it still computes the ANOVA, and Levene’s test will flag the unequal spread.
Levene’s test needs the raw values, so it appears in Raw data mode only.
Worked examples
Example 1 — raw data (the calculator’s default). Three groups: A = 6, 8, 7, 9, 8; B = 5, 4, 6, 5, 7; C = 9, 10, 8, 9, 12. Group means are 7.6000, 5.4000 and 9.6000; the grand mean is 7.5333. SSB = 44.1333, SSW = 19.6000 and SST = 63.7333. With df (2, 12), MSB = 22.0667 and MSW = 1.6333, so F = 13.5102 and p = 0.0008 — reject H₀ at the 5% level. η² = 0.6925, ω² = 0.6252 and Cohen’s f = 1.5006. Levene’s test gives F = 0.0690 with p = 0.9337, so the equal-variance assumption holds. Tukey HSD (q = 3.7729, HSD = 2.1564): A−B is significant (2.2000), A−C is not (−2.0000) and B−C is significant (−4.2000).
Example 2 — summary stats. Three groups of 10: means 72, 78 and 84, with SDs 6, 7 and 5. The grand mean is 78.0000. SSB = 720.0000, SSW = 990.0000, so with df (2, 27), MSB = 360.0000 and MSW = 36.6667, giving F = 9.8182 and p = 0.0006 — reject H₀. η² = 0.4211 and ω² = 0.3702. Tukey HSD and Levene’s test need the raw values, so they do not appear in this mode.
Example 3 — unequal group sizes (Tukey–Kramer). A = 12, 14, 13, 15; B = 10, 9, 11, 10, 12; C = 16, 18, 17, 19, 15, 17. Group means are 13.5000, 10.4000 and 17.0000. SSB = 119.5333 and SSW = 20.2000, so with df (2, 12), F = 35.5050 and p < 0.0001 — strongly significant. η² = 0.8554. The Tukey–Kramer HSD for A−B is 2.3220 (q = 3.7729, MSW = 1.6833).
All values on this page were checked against scipy.stats.f_oneway.
Frequently asked questions
When do I use ANOVA instead of a t-test?
Use ANOVA when you have three or more groups to compare. A t-test compares two means; with three or more groups, running several t-tests inflates the chance of a false positive, while ANOVA tests all the means at once with a single error rate. If you have exactly two groups, use a two-sample t-test — it gives the same answer as ANOVA (F = t²). See ANOVA vs t-test for the full comparison.
What do eta squared and omega squared mean?
Eta squared (η²) is the share of the total variation that lies between the groups: η² = SSB ÷ SST. Omega squared (ω²) is the same idea with a correction for bias, so it estimates the population effect more honestly and is always a little smaller. Roughly, 0.01 is a small effect, 0.06 is medium and 0.14 is large. Both are reported alongside the p-value because a significant p-value alone does not say how big the difference is.
What does Tukey HSD do after a significant ANOVA?
A significant ANOVA says at least one group mean differs, but not which ones. Tukey’s HSD compares every pair of groups and flags the pairs whose means differ by more than the HSD — a threshold built from the studentised range critical value q, the within-groups mean square and the group sizes. It holds the overall error rate at α across all the comparisons, so it is safer than running separate t-tests on every pair. With unequal group sizes it uses the Tukey–Kramer form. It needs the raw values.
What if Levene’s test rejects equal variances?
Then the equal-variance assumption behind the ordinary ANOVA is in doubt, and its p-value may be misleading. The standard remedies are Welch’s ANOVA, which does not assume equal variances, or the Kruskal–Wallis test, a rank-based alternative that does not assume normality either. Both are outside this calculator; the usual advice is to run one of them and compare its conclusion with the ANOVA result. A single group with zero variance (all values identical) does not stop this calculator, but Levene’s test will flag it.
Source: the one-way ANOVA table, Tukey HSD and Levene’s test follow the standard reference for analysis of variance (Fisher’s ANOVA). All examples on this page were checked against scipy.stats.f_oneway.
See all 67 free statistics tools on Empirical Rule Calculators.
Next steps
🧮
● Live
F-Distribution Calculator
Look up F critical values and run the F-test for two variances.
Calculator
Try Calculator →
🧮
● Live
Two-Sample T-Test Calculator
Compare two means — for two groups it agrees with ANOVA (F = t²).
Calculator
Try Calculator →
📘
● Live
ANOVA vs t-Test: When to Use Each One
When to run ANOVA and when a t-test is enough.
Article
Read Article →
More Tools to Explore
Keep going with the calculators that pair with one-way ANOVA—the F-distribution, the two-sample t-test, and the chi-square test.
F-Distribution Calculator
Look up any F probability or critical value, and run the F-test for two variances.
Two-Sample T-Test Calculator
Compare two means—for two groups it agrees with ANOVA (F = t²).
ANOVA vs t-Test: When to Use Each One
When to run ANOVA and when a t-test is enough.
Chi-Square Test Calculator
Test counts and categories with the chi-square test.
P-Value Calculator from Z-Score
Turn any z-score into a p-value, one- or two-sided.
Explore All Tools→
Browse all 67 free statistics tools—every calculator on this site in one place.