How this instrument works
Analysis of variance (ANOVA) answers a question a series of separate two-group comparisons can't cleanly answer: do three or more groups share the same underlying mean, or does at least one differ? Running three separate two-sample tests on three groups (A vs B, B vs C, A vs C) inflates the chance of a false alarm; one-way ANOVA instead tests all three groups in a single pass by comparing how much the group means vary against each other to how much the individual observations vary within each group.
The comparison happens through a sum-of-squares breakdown. The grand mean is the average of every observation across all three groups pooled together. The sum of squares between groups (SSB) measures how far each group's own mean sits from that grand mean, weighted by group size — a large SSB means the groups' averages are spread apart. The sum of squares within groups (SSW) measures how much observations scatter around their own group's mean — a large SSW means there's a lot of natural noise inside each group regardless of any real difference between them.
Dividing each sum of squares by its degrees of freedom gives a mean square, and the F-statistic is simply the ratio of the between-groups mean square to the within-groups mean square. A large F means the groups' means are spread further apart than the within-group noise alone would predict — evidence that at least one group's true mean differs from the others. A large F does not say which group differs, or by how much; that follow-up question needs a post-hoc comparison, which this instrument does not perform.
- Enter the raw values for the first group into Group 1, separated by commas, spaces, or new lines.
- Enter the raw values for the second and third groups into Group 2 and Group 3 the same way.
- Each group needs at least 2 values so a within-group spread can be computed.
- Read Grand mean, Sum of squares between groups (SSB), and Sum of squares within groups (SSW) for the underlying breakdown.
- Read the F-statistic (MSB / MSW) — a large value relative to an F-distribution critical point (using dfB and dfW at your chosen significance level) signals that the three groups' means likely aren't all equal.
Worked example — three study-method groups' quiz scores
Nine students are split into three study-method groups of three and take the same quiz: Group 1 scores 2, 3, 4; Group 2 scores 5, 6, 7; Group 3 scores 8, 9, 10 (all out of 10). Each group's own mean is 3, 6, and 9, and each group has an identical internal spread — variance 1 in every group, since each set of three scores deviates by exactly −1, 0, +1 from its own mean.
The grand mean across all nine scores is (3×3 + 3×6 + 3×9)/9 = 54/9 = 6.0. Sum of squares between groups (SSB) is 3(3−6)² + 3(6−6)² + 3(9−6)² = 27+0+27 = 54.0, and Sum of squares within groups (SSW) is 2×1 + 2×1 + 2×1 = 6.0. With dfB=2 and dfW=6, the mean squares are 54/2=27 and 6/6=1, so the F-statistic (MSB / MSW) reads 27.0 — a large ratio driven entirely by how cleanly separated the three groups' means are relative to their (identical, modest) internal spread.
Questions
What does a large F-statistic actually tell you?
It tells you the spread between the three groups' means is large relative to the ordinary scatter inside each group — evidence against the hypothesis that all three groups share one true mean. On its own, a large F doesn't identify which group (or groups) differ from the others; a post-hoc test such as Tukey's HSD is the standard next step for pinpointing that, and this instrument doesn't run one.
Why does this calculator require exactly three groups?
It's built specifically for the common three-group case, taking each group's raw data as a separate input. The underlying one-way ANOVA method generalizes to any number of groups (two or more), but this instrument's inputs and sum-of-squares formulas are fixed to three.
What should I do after getting the F-statistic?
Compare it against a critical value from the F-distribution using dfB and dfW (reported here) at the significance level you've chosen — most statistics software or an F-table can supply that critical value. If F exceeds it, you have evidence that not all three group means are equal; if it doesn't, you don't have enough evidence to say they differ.
Does ANOVA assume the three groups have similar variances?
Yes — one-way ANOVA assumes roughly equal variances across groups (homogeneity of variance) along with independent, roughly normally distributed observations within each group. When variances differ sharply between groups, the F-statistic becomes less reliable, and a variance-robust alternative such as Welch's ANOVA is usually preferred instead.
Can I use this instrument to compare just two groups?
It requires all three group fields to be filled, so it isn't set up for a pure two-group comparison — a two-sample t-test is the standard, more direct tool when you only have two groups to compare. One-way ANOVA on two groups produces an F-statistic mathematically equivalent to the square of that t-test's t-statistic, but this instrument's inputs assume three groups are genuinely being compared.