ANOVA Calculator

How to Use This Calculator

1. Enter raw values for each group as comma-separated numbers in the fields.
2. The calculator computes F-statistic, SS between/within, and degrees of freedom.
3. Switch to Two-Way ANOVA tab to input SS and df directly for factorial designs.
4. Use Repeated Measures tab for within-subjects designs.
5. Professional tab adds η² effect size, Tukey HSD, and a full ANOVA table.

Formula

Example

Frequently Asked Questions

• What is ANOVA and how does it differ from t-tests? ▼
  ANOVA (Analysis of Variance) is a statistical procedure that tests whether the means of three or more groups are equal by comparing variance between groups to variance within groups. The resulting F-statistic is the ratio of mean square between groups to mean square within groups. When that ratio is large, the between-group variation exceeds what random chance would produce, suggesting at least one group mean differs. The key distinction from t-tests is scale: a t-test compares exactly two means, while ANOVA handles any number of groups simultaneously. More importantly, running multiple t-tests inflates the Type I error rate (false positives). With three groups, three separate t-tests each at α=0.05 give a family-wise error rate of roughly 14% instead of 5%. ANOVA controls this by performing a single omnibus test. ANOVA was developed by R.A. Fisher in the 1920s and remains the backbone of experimental design across medicine, psychology, agriculture, and engineering.
• When should I use one-way vs two-way ANOVA? ▼
  One-way ANOVA tests the effect of a single categorical independent variable (factor) on a continuous dependent variable. Example: comparing exam scores across three teaching methods. Two-way ANOVA tests two independent factors simultaneously and, crucially, their interaction. Example: examining how exam scores are affected by both teaching method and student gender, plus whether gender moderates the effect of teaching method (the interaction term). Two-way ANOVA is more efficient than running two separate one-way ANOVAs because it uses all the data at once and explicitly estimates the interaction. If you have only one factor of interest, use one-way ANOVA. If you suspect that two factors jointly influence the outcome — or if you want to control for a second variable to increase precision — use two-way ANOVA. For more than two factors, use n-way ANOVA or a mixed-effects model, though interpretation of high-order interactions becomes challenging.
• What's a post-hoc test and why do I need it after ANOVA? ▼
  ANOVA's F-test is an omnibus test — it tells you that at least one group mean differs from the others, but not which pairs differ. A post-hoc test performs all pairwise comparisons while controlling the family-wise error rate. Tukey's Honestly Significant Difference (HSD) is the most common: it computes the minimum difference between two group means needed to declare significance, based on the Studentized range distribution. Tukey is conservative and appropriate when comparing all possible pairs. Bonferroni correction is simpler — it divides α by the number of comparisons — but is more conservative than Tukey for large numbers of groups. Scheffé's test is the most conservative and valid even for unplanned comparisons. Dunnett's test is optimal when comparing multiple groups to a single control group. Always choose your post-hoc test before seeing the data to avoid selective reporting. Running post-hoc tests only when ANOVA is significant is correct; running them without ANOVA is generally discouraged.
• How do I interpret eta-squared from ANOVA results? ▼
  Eta-squared (η²) is ANOVA's primary effect size measure, defined as SS_between / SS_total. It represents the proportion of total variance in the dependent variable accounted for by the group membership factor. Jacob Cohen's conventional benchmarks: η² < 0.01 is negligible, 0.01–0.06 is small, 0.06–0.14 is medium, and ≥ 0.14 is large. A statistically significant F-test with η² = 0.02 indicates a real but practically small effect — group membership explains only 2% of outcome variance. Conversely, a non-significant test with η² = 0.12 in a small underpowered study might represent a meaningful effect that deserves further investigation. Partial eta-squared (η²_p) is used in two-way and higher ANOVA: it divides SS_effect by (SS_effect + SS_error), excluding other factors from the denominator. Omega-squared (ω²) is a less biased estimator of population effect size and is preferred for reporting in journal articles.
• What assumptions must be met for ANOVA to be valid? ▼
  ANOVA rests on three main assumptions. First, normality: the dependent variable should be approximately normally distributed within each group. ANOVA is fairly robust to mild non-normality, especially with equal sample sizes and n > 15 per group. Shapiro-Wilk or Kolmogorov-Smirnov tests can check this; Q-Q plots are more informative. Second, homogeneity of variance (homoscedasticity): group variances should be approximately equal. Levene's test formally checks this. Welch's ANOVA is the robust alternative when variances are unequal. Third, independence: observations must be independent of each other — no repeated measures, clustering, or nesting (for standard one-way ANOVA). When assumptions are violated: use Kruskal-Wallis (non-parametric one-way ANOVA) if normality fails, Welch's ANOVA if variances are unequal, and repeated measures ANOVA or mixed models if observations are correlated. Transforming the dependent variable (log, square root) can sometimes restore normality and/or variance equality.

Related Calculators