ANOVA Calculator
Compare three or more group means — one-way, two-way, or with replication. Get the F-statistic, p-value, and post-hoc pairwise comparisons. Runs in your browser.
Open the calculator →What is ANOVA, and when do you need it?
You've got three groups — say, a low-dose drug, a high-dose drug, and a placebo. You want to know if any group mean is different from the others. You could run three separate t-tests. But here's the problem: each test has a 5% false positive rate. With three tests, your actual chance of at least one false alarm goes up to 14.3%. Six groups? That's 54%.
ANOVA solves this with a single omnibus test that keeps your false positive rate at exactly 5%. If it's significant, then you do pairwise comparisons.
Why not just run multiple t-tests?
I hear this question all the time. "Why can't I just do A vs. B, A vs. C, B vs. C?" The math is simple: if each test has a 5% chance of a false positive, then with 3 tests:
1 − (0.95)³ = 14.3% false positive rate
With 6 groups, you'd have 15 pairwise comparisons. Your false positive rate? 54%. That's worse than a coin flip. ANOVA keeps it at 5% because it does one test across all groups.
How to use the calculator
- Paste your data. One group per line, values separated by spaces or commas.
- Hit Calculate. You'll get the F-statistic, p-value, and eta-squared.
- Check p < 0.05? If yes, at least two groups differ. Now you need post-hoc tests to figure out which ones.
- Look at eta-squared. This tells you what proportion of variance the groups explain. F-statistic alone doesn't tell you if the difference matters in practice.
A real example: does dosage matter?
Three dosage levels of an enzyme inhibitor. Six samples each:
10mg: 45, 48, 42, 47, 44, 46
25mg: 52, 55, 50, 54, 51, 53
50mg: 58, 62, 56, 60, 57, 59
F(2,15) = 28.4, p < 0.0001, η² = 0.79 (large effect)
There is a significant difference between at least two dosage groups. The large eta-squared (0.79) means 79% of the variance in enzyme activity is explained by dosage level. Post-hoc Tukey HSD reveals all three pairs differ significantly (p < 0.01 for each).
Effect Size: Eta-Squared (η²)
| η² Value | Interpretation | Practical Meaning |
|---|---|---|
| 0.01 | Small | 1% of variance explained by group |
| 0.06 | Medium | 6% of variance explained |
| 0.14 | Large | 14%+ of variance explained |
Assumptions to Check
- Normality: Data in each group should be approximately normally distributed. Check with Shapiro-Wilk test. ANOVA is robust to mild violations when group sizes are similar.
- Homogeneity of variance: Groups should have similar variances. Check with Levene's test. If violated, use Welch's ANOVA.
- Independence: Observations should be independent of each other.
- No outliers: Extreme values can dramatically inflate or deflate the F-statistic.
When ANOVA Is Significant: Post-Hoc Tests
A significant ANOVA tells you "at least two groups differ" but not which ones. You need post-hoc tests:
- Tukey HSD: Most common. Controls family-wise error rate for all pairwise comparisons. Best when group sizes are equal.
- Bonferroni: Conservative. Adjusts p-values for multiple comparisons. Good for small number of comparisons.
- Holm: Step-down procedure. Less conservative than Bonferroni, more powerful.
- Scheffe: Most conservative. Good for complex contrasts and unequal group sizes.
What usually goes wrong
- Running t-tests instead of ANOVA. 5% → 14% false positive rate with just 3 groups. Use ANOVA, then do post-hoc pairwise comparisons.
- Skipping assumption checks. Non-normal data with small samples gives you garbage p-values. Run Shapiro-Wilk before anything else.
- Not reporting effect size. F = 4.2, p = 0.03 means nothing if η² = 0.02. That's 2% of variance explained. Statistically significant and practically pointless.
- Stopping at the omnibus test. "ANOVA was significant" doesn't tell you which groups differ. Tukey HSD will — run it.
Try the ANOVA Calculator
Paste one group per line, click Calculate, and get F-statistic, p-value, eta-squared, and post-hoc comparisons. No login required.
Open the Calculator →