One-Way ANOVA Calculator
Paste one column per group, with the group names on top…
Paste one column per group, with the group names on top, to test whether any of the means differ with a one-way analysis of variance.
Read 24 values in 3 groups; left out the header row (“Red”, “Blue”, “White”).
More options95% confidence
The significance level is 1 − the confidence level (0.05 at 95%).
ExampleP-value
1.13 × 10⁻⁷
At least one group’s mean differs from the others: Blue is highest at 17.188, Red lowest at 14.425. Significant at the 5% level (p < 0.05).
Paste your own numbers to test them.
Every value is a dot and the box is the middle half; the diamond is the mean with its confidence interval.
The F distribution if every group had the same mean; the shaded tail beyond the observed F is the p-value.
- F, df
- 37.6808, (2, 21)
- η² (share of variation)
- 0.782 (ω² 0.753)
- Groups · values
- 3 · 24
How this works: 9 sources,
How this works
Method
One-way analysis of variance tests whether several groups share the same mean, in one test instead of a t-test for every pair. Running three t-tests on three groups at 5% each raises the chance of at least one false alarm to about 14%; ANOVA keeps it at 5%. In the example, seedlings grown under blue light averaged 17.2 cm against 14.4 cm under red and 15.6 cm under white. Full method
Sources
- NIST/SEMATECH e-Handbook of Statistical Methods, 7.4.3 Are the means equal? (one-way ANOVA)
- Wikipedia, One-way analysis of variance
- NIST/SEMATECH e-Handbook of Statistical Methods
- Abramowitz and Stegun, Handbook of Mathematical Functions, chapters 6, 7 and 26 (1964)
- SciPy 1.18 scipy.stats (reference values for the tests)
- statsmodels 0.15 (reference values for the A/B test and regression intervals)
- Welch, The generalization of Student’s problem when several different population variances are involved (Biometrika) (1947)
- Fleiss, Levin and Paik, Statistical Methods for Rates and Proportions, 3rd edition (sample size for two proportions) (2003)
- Hyndman and Fan, Sample quantiles in statistical packages (The American Statistician) (1996)
How it’s tested
2 worked examples for this page are checked by automated tests before every release: given the inputs, the tool must show the expected answer.
Privacy
On your device. Nothing you enter or open leaves this browser. About privacy labels
Changes
- First published: live curves with draggable cut-offs, and tests on pasted data (t-tests, ANOVA, chi-square, Fisher’s exact, correlation, regression, A/B), checked against SciPy.
How one-way ANOVA works
One-way analysis of variance tests whether several groups share the same mean, in one test instead of a t-test for every pair. Running three t-tests on three groups at 5% each raises the chance of at least one false alarm to about 14%; ANOVA keeps it at 5%. In the example, seedlings grown under blue light averaged 17.2 cm against 14.4 cm under red and 15.6 cm under white.
The idea is in the name: it splits the total variation into the part between the group means and the part within the groups. F is the ratio of the two mean squares, (SS between ÷ (k − 1)) ÷ (SS within ÷ (N − k)). If the groups were all alike, F would hover around 1; here it is about 38, because the light colours pull the means far apart compared with how much the plants vary under the same light.
A small p-value says at least one mean differs, not which. Look at the plot, then compare pairs with a correction for multiple tests (Tukey’s HSD is the usual one) if you need to name them. η² (eta squared) is the share of all the variation explained by the groups; ω² is a less biased version of it.
F = [Σ nᵢ(x̄ᵢ − x̄)² ÷ (k − 1)] ÷ [Σ Σ (x − x̄ᵢ)² ÷ (N − k)]
- One-way ANOVA compares k group means by the ratio of the between-group mean square to the within-group mean square, which follows an F distribution with k − 1 and N − k degrees of freedom when the means are equal. Source: NIST/SEMATECH e-Handbook of Statistical Methods, 7.4.3 Are the means equal? (one-way ANOVA).
- ANOVA assumes the groups are independent, roughly normal and have similar variances; it is fairly robust to non-normality with equal group sizes. Source: Wikipedia, One-way analysis of variance.
The F value needed for significance
Critical F at the 5% and 1% levels for k groups of n values each (degrees of freedom k − 1 and k(n − 1)).
| Groups × per group | df | F (5%) | F (1%) |
|---|---|---|---|
| 2 × 5 | 1, 8 | 5.318 | 11.259 |
| 3 × 5 | 2, 12 | 3.885 | 6.927 |
| 3 × 8 | 2, 21 | 3.467 | 5.78 |
| 3 × 10 | 2, 27 | 3.354 | 5.488 |
| 3 × 20 | 2, 57 | 3.159 | 4.998 |
| 4 × 5 | 3, 16 | 3.239 | 5.292 |
| 4 × 10 | 3, 36 | 2.866 | 4.377 |
| 4 × 25 | 3, 96 | 2.699 | 3.992 |
| 5 × 6 | 4, 25 | 2.759 | 4.177 |
| 5 × 10 | 4, 45 | 2.579 | 3.767 |
Show all 12 rowsShow fewer
| 6 × 10 | 5, 54 | 2.386 | 3.377 |
| 8 × 10 | 7, 72 | 2.14 | 2.898 |
Common questions
How do I paste the groups?
One column per group, with the group names in the first row, straight from a spreadsheet. Columns can be different lengths; blank cells at the bottom are fine.
ANOVA says there is a difference. Which groups differ?
The F test doesn’t say. Look at the means and their intervals in the plot, then run a post-hoc test such as Tukey’s HSD, or t-tests with a Bonferroni correction (multiply each p-value by the number of comparisons).
Is a one-way ANOVA with two groups the same as a t-test?
Yes, for Student’s t-test: F equals t² and the p-values match. With unequal variances, Welch’s t-test is the safer choice for two groups.
What if my groups have very different spreads?
The classic F test assumes similar variances. With equal group sizes it copes with moderate differences; with unequal sizes, consider Welch’s ANOVA, or the Kruskal-Wallis rank test for skewed data.