Compare three or more independent groups with a one-way ANOVA calculator that returns the F statistic, p-value, ANOVA table, group means, effect sizes.
Last updated
Hypothesis testing
ANOVA calculator
Compare three or more independent groups with a one-way analysis of variance. Enter raw values for each group and the calculator returns the F statistic, p-value, ANOVA table, group summaries, and effect-size context so you can see whether the group means differ beyond random variation.
Quick examples
Groups
Each group should contain at least two numeric values. Balanced groups are ideal, but unequal sample sizes are allowed.
Significance level
How to read the result
Use raw values from independent groups, not percentages or matched pairs.
If the p-value is below α, at least one group mean is different.
After a significant result, follow up with a post-hoc test such as Tukey HSD to locate the difference.
Result
F = 141.56
The one-way ANOVA suggests that at least one group mean differs from the others (F(2, 12) = 141.56, p < .001). The largest mean is Treatment B and the smallest mean is Control; use the pairwise guide below as a screen, then confirm with a planned post-hoc test such as Tukey HSD or Games-Howell when reporting formal results.
141.56
F statistic
< .001
p-value
0.96
η² effect size
0.95
ω² effect size
Reject H₀
Decision at α = 0.05
1.54
Largest mean gap
1.92
Largest / smallest variance
Balanced
Sample balance
Key ANOVA totals
Between groups SS
5.95
Within groups SS
0.25
Total SS
6.2
df between / within / total
2 / 12 / 14
MS between
2.97
MS within
0.02
ANOVA source table
Source
SS
df
MS
F
p
Between groups
5.95
2
2.97
141.56
< .001
Within groups
0.25
12
0.02
—
—
Total
6.2
14
—
—
—
Group summary
Group
n
Mean
SD
Variance
Control
5
4.06
0.11
0.01
Treatment A
5
4.9
0.16
0.02
Treatment B
5
5.6
0.16
0.02
Pairwise mean-gap guide
Exploratory pairwise checks using the ANOVA pooled error term and a Bonferroni-adjusted p-value. Use this as a screen before reporting a planned post-hoc test.
Pair
Mean gap
SE
t guide
Bonferroni p
Screen at 0.05
Control vs Treatment B
-1.54
0.09
-16.8
< .001
Likely gap
Control vs Treatment A
-0.84
0.09
-9.17
< .001
Likely gap
Treatment A vs Treatment B
-0.7
0.09
-7.64
< .001
Likely gap
Reject H₀ at α = 0.05
Assumptions and follow-up
Use raw numeric observations from independent groups rather than percentages or repeated measures.
Each group should be approximately normal, especially when sample sizes are small.
Group variances should be broadly similar; if they are not, Welch's ANOVA is usually the safer alternative.
A significant result tells you that at least one group mean differs, not which groups differ.
ANOVA calculator: compare three or more group means with an F test
A one-way ANOVA calculator compares the means of three or more independent groups and tells you whether the differences are large enough to matter statistically. It returns the F statistic, p-value, ANOVA table, group summaries, and effect-size context from raw data, which makes it a practical companion when you want a single test instead of several pairwise t-tests.
What one-way ANOVA tests
ANOVA stands for analysis of variance. In a one-way ANOVA, the question is whether the mean of one group is different from the others when you have one independent variable and three or more groups. The null hypothesis says all population means are equal; the alternative says at least one group mean is different.
This is the right tool when you have independent samples and want to compare more than two means at the same time. If you only have two groups, a t-test is the simpler comparison. If your measurements are paired or repeated on the same subjects, a repeated-measures approach is the better fit.
How the F ratio is built
One-way ANOVA separates the total spread in the data into between-group variation and within-group variation. The between-group sum of squares measures how far each group mean sits from the grand mean, while the within-group sum of squares measures how spread out the values are inside each group.
The test statistic is the F ratio: F = MS_between / MS_within. If the group means are very similar compared with the spread inside the groups, the F value stays close to 1. If the group means are much farther apart than the within-group variation, the F value grows and the right-tailed p-value shrinks.
The same decomposition also appears in the ANOVA table, which lists the source of variation, sums of squares, degrees of freedom, and mean squares. That table is what lets you check the arithmetic instead of treating the output as a black box.
SS_between = Σ n_i( x̄_i − x̄ )²
Between-group sum of squares, using each group mean and the grand mean.
SS_within = ΣΣ (x_ij − x̄_i)²
Within-group sum of squares, using each observation around its own group mean.
F = MS_between / MS_within
The ANOVA F ratio compares explained variation to residual variation.
η² = SS_between / SS_total
Eta-squared effect size, which shows how much of the total variation is associated with group membership.
Worked example
Take the default example on the page: three groups labelled Control, Treatment A, and Treatment B. The Control values cluster around 4, Treatment A values cluster around 5, and Treatment B values cluster around 5.5. That separation produces a very large F statistic and a very small p-value, which is exactly what you would expect when the group means are clearly different relative to the variation inside each group.
The helpful part of the result sheet is not just the headline F test. The group-summary table shows the mean and standard deviation for each group, while the ANOVA table shows how much variation sits between groups and how much stays within groups. That makes it easier to explain why the result is significant instead of simply reporting that it is.
Reading the result
If the p-value is below your chosen alpha level, usually 0.05, you reject the null hypothesis and conclude that at least one group mean differs. If the p-value is above alpha, the data do not provide strong evidence against equal means.
A significant ANOVA does not tell you which groups differ. It only tells you that the means are not all the same. If you need pairwise detail, a post-hoc test such as Tukey HSD is the usual follow-up because it controls the family-wise error rate across multiple comparisons.
Effect size adds useful context. Eta-squared tells you the proportion of total variation associated with group membership, while omega-squared gives a slightly more conservative estimate of practical impact. Those values do not replace the p-value, but they help you decide whether the difference looks meaningful as well as statistically detectable.
Using the pairwise mean-gap guide
Many ANOVA calculator searches do not stop at the omnibus F test. People also want to know which groups might be driving the result. The calculator now adds an exploratory pairwise mean-gap guide that sorts every group pair by the size of the mean difference, shows the standard error from the ANOVA pooled error term, and applies a simple Bonferroni p-value screen.
That guide is not a replacement for a planned post-hoc test in a formal report. Treat it as a fast interpretation layer: it points you toward the largest candidate gaps, while the assumptions panel tells you whether Tukey HSD, Games-Howell, Welch's ANOVA, or a non-parametric alternative may be more appropriate before you publish or submit the analysis.
Assumptions and alternatives
A standard one-way ANOVA assumes independent observations, approximately normal group distributions, and broadly similar variances. The test is fairly robust to mild violations, especially with balanced and moderately sized samples, but large variance differences can distort the result.
If the groups have noticeably unequal variances, Welch's ANOVA is often the safer alternative. If the data are strongly non-normal or ordinal, a non-parametric test such as Kruskal-Wallis may be more appropriate. Those alternatives answer a similar comparison question, but they relax different assumptions.
Unequal sample sizes are allowed, but balanced groups are easier to interpret and usually give a more stable estimate of the F ratio. That is why many ANOVA examples keep the groups even when the maths itself does not require it.
This page is a calculation aid for introductory statistics, teaching, and research planning. It does not run post-hoc tests, and it does not replace software for more advanced factorial or repeated-measures designs.
For a deeper reference on one-way ANOVA calculations and the ANOVA table, see the NIST/SEMATECH handbook and the OpenStax one-way ANOVA chapter linked below.
You need at least three groups for a one-way ANOVA, because the point of the test is to compare three or more means at the same time. If you only have two groups, a t-test is simpler and gives the same basic comparison more directly.
What does a significant ANOVA mean?
A significant ANOVA means the F test found enough evidence to say the group means are not all the same. It does not tell you which groups differ from one another; it only tells you that at least one mean is different. A post-hoc test such as Tukey HSD is the usual next step when you need that detail.
Can ANOVA handle unequal sample sizes?
Yes. One-way ANOVA does not require identical sample sizes in each group. Balanced samples are usually easier to interpret and can give a more stable estimate, but unequal group sizes are common in real data and are still valid as long as the other assumptions are reasonable.
What assumptions does one-way ANOVA make?
One-way ANOVA assumes independent observations, approximately normal group distributions, and broadly similar variances across groups. The test is fairly robust to mild assumption violations, but large departures from normality or very uneven variances can make the p-value less trustworthy.
When should I use Welch's ANOVA instead?
Use Welch's ANOVA when the groups have clearly different variances or the sample sizes are unbalanced enough that the equal-variance assumption looks weak. Welch's version is designed to be more robust in that situation and is often a better default when the groups are not nicely matched.
Do I need a post-hoc test after ANOVA?
You need a post-hoc test when the overall ANOVA is significant and you want to know which specific group pairs differ. Common choices are Tukey HSD and Games-Howell, depending on whether the variances are similar. If the overall ANOVA is not significant, post-hoc testing usually does not add much.
Is the pairwise table the same as Tukey HSD?
No. The pairwise table is an exploratory screening guide based on the ANOVA pooled error term and a Bonferroni-adjusted p-value. Tukey HSD and Games-Howell are formal post-hoc procedures with their own assumptions. Use the table to see which gaps deserve attention, then use the appropriate planned post-hoc method when you need report-ready pairwise conclusions.
Can I use ANOVA for repeated measures or paired data?
Not with this calculator. One-way ANOVA assumes independent groups, so repeated measurements on the same subjects or matched pairs need a repeated-measures or paired analysis instead. Treating linked observations as independent would understate the true error and can distort the p-value.
