One-Way ANOVA Calculator

How to Use the One-Way ANOVA Calculator

1. 1. Enter each group. Name the group and paste its scores, separated by commas, spaces, or new lines. Three groups are ready to fill; add up to ten or remove down to two.
2. 2. Choose α. The significance level (.05 by default) drives the decision line and the verdicts on the follow-up tests.
3. 3. Read the omnibus test. The ANOVA table gives SS, df, MS, F, and p; η² and ω² say how much of the variation the groups explain.
4. 4. Check and follow up. Look at Levene's test before trusting the classic F, use Welch's ANOVA if variances differ, then read the pairwise comparisons to see which groups differ. Copy the APA line when you are done.

What the F Ratio Compares

ANOVA splits the total variation in the scores into two parts. The between-groups sum of squares, SS_B = Σ nⱼ(x̄ⱼ − x̄)², measures how far each group mean sits from the grand mean. The within-groups sum of squares, SS_W = Σ Σ (x − x̄ⱼ)², measures how far individual scores sit from their own group's mean. Dividing each by its degrees of freedom gives a mean square, and F = MS_B / MS_W. If the population means are equal, both mean squares estimate the same error variance and F hovers around 1; a large F means the group means are further apart than the noise inside the groups would produce by chance.

Worked Example: Three Teaching Formats

Click Try an example to load quiz scores from three classes of six students: lecture (M = 73.83, SD = 4.36), flipped (M = 82.17, SD = 3.76), and self-paced online (M = 72.33, SD = 2.58). The grand mean is 76.11.

In APA style: quiz scores differed by teaching format, F(2, 15) = 12.69, p < .001, η² = .63. Levene's test gives no sign of unequal variances, F(2, 15) = 0.89, p = .430, and Welch's ANOVA agrees with the classic test, F(2, 9.49) = 13.33, p = .002. The Bonferroni-adjusted comparisons place the difference in the flipped class: it outscored lecture by 8.33 points (adjusted p = .016) and online by 9.83 points (adjusted p = .002), while lecture and online did not differ (adjusted p > .999).

Assumptions and What to Do When They Fail

• Independent observations. Each score comes from a different person and belongs to one group. Repeated measurements on the same people need a repeated-measures ANOVA, which this calculator does not run.
• Roughly normal residuals. ANOVA tolerates moderate skew when groups are similar in size; with small, clearly non-normal groups, use the Kruskal-Wallis test.
• Equal variances. Levene's test here is the Brown–Forsythe version, which measures spread around each group's median and stays reliable with skewed data. When it is significant — or when group sizes differ a lot — report Welch's ANOVA, which drops the assumption.
• Post hoc tests only after a reason. The pairwise table uses Welch t-tests with a Bonferroni adjustment. Tukey's HSD is not computed because it needs the studentized range distribution; Bonferroni is slightly more conservative, so treat borderline adjusted p-values with care.

Frequently Asked Questions

What does a significant one-way ANOVA tell you?

That at least one group mean differs from the others — not which one, and not that all of them differ. The omnibus F test only compares the spread of the group means with the spread inside the groups. To locate the difference, read pairwise comparisons such as the Bonferroni-adjusted tests shown with the results.

Why not just run several t-tests?

Each t-test carries its own chance of a false positive. With three groups there are three pairwise tests, and at α = .05 each, the chance of at least one false alarm climbs toward 1 − .95³ ≈ 14%. ANOVA tests all the means at once at the chosen α, and adjusted post hoc tests keep the family of comparisons under control.

What is the difference between η² and ω²?

Both describe the proportion of total variance explained by group membership. η² = SS_B / SS_T is the sample proportion and overestimates the population value, especially with small samples. ω² subtracts the variance you would expect by chance and is the less biased estimate, so it is always a little smaller. Report η² if your field expects it, ω² when samples are small.

When should I use Welch's ANOVA instead of the classic F test?

When the group variances differ — especially if the group sizes differ too. The classic F test pools the variances, and if a small group has the largest variance its false-positive rate can climb well above α. Welch's ANOVA weights each group by n / s² and adjusts the error degrees of freedom, so it stays accurate; with equal variances it gives almost the same answer.

How do I report a one-way ANOVA in APA style?

Give the group means and SDs, then F(df between, df within) = value, p = value, and an effect size: F(2, 15) = 12.69, p < .001, η² = .63. Italicise F and p but not the Greek η; drop the leading zero from p and η² because neither can exceed 1. Follow with the post hoc results that show which groups differ.

How many scores do I need in each group?

The calculator needs at least two per group, but that is a computational floor, not a design target. To detect a medium effect (f = .25) with three groups at α = .05 and 80% power you need about 53 per group. Equal group sizes also make the test more robust to unequal variances.