What is ANOVA?
ANOVA (Analysis of Variance) is a statistical technique used to test whether the means of two or more groups are significantly different from one another. Instead of running multiple t-tests, which inflates the risk of false positives, ANOVA evaluates all groups simultaneously with a single F-test.
The core idea is simple: if the variation between groups is much larger than the variation within groups, that is evidence the group means are not equal. ANOVA quantifies this ratio as the F-statistic.
**When to use ANOVA:** - You have three or more independent groups to compare - Your outcome variable is continuous (e.g., test scores, blood pressure, yield) - You want to know whether any group differs significantly from the others
For two groups, a two-sample t-test is equivalent and often preferred. For three or more groups, ANOVA is the standard approach.
One-Way vs Two-Way ANOVA
**One-Way ANOVA** examines how a single categorical factor (the grouping variable) affects a continuous outcome. For example: does fertilizer type (A, B, C) affect crop yield?
**Two-Way ANOVA** examines two categorical factors simultaneously and their interaction. For example: does fertilizer type and watering frequency jointly affect crop yield, and do those two factors interact?
This calculator focuses on **One-Way ANOVA**, which is the most common form and the correct choice when you have a single grouping variable with three or more levels.
The F-Statistic and p-Value
The F-statistic is the ratio of variance between groups to variance within groups:
\`\`\` F = MS_between / MS_within
MS_between = SS_between / df_between MS_within = SS_within / df_within
df_between = k - 1 (k = number of groups) df_within = N - k (N = total observations) \`\`\`
A large F-statistic means the between-group spread is much larger than the within-group spread, suggesting the group means are genuinely different.
The **p-value** is the probability of observing an F-statistic at least as large as the one you got, assuming the null hypothesis (all means are equal) is true. If p < 0.05, you reject the null hypothesis at the conventional 5% significance level.
**Effect size (eta-squared, η²)** measures how much of the total variance is explained by the group factor: \`\`\` η² = SS_between / SS_total \`\`\` Conventions: η² < 0.06 = small, 0.06–0.14 = medium, > 0.14 = large.
Post-Hoc Tests: Tukey HSD
When ANOVA is significant, you know at least one group differs — but not which pairs. Post-hoc tests make pairwise comparisons while controlling the family-wise error rate.
**Tukey's Honest Significant Difference (HSD)** is the most widely used post-hoc test. For each pair of groups, it computes a q statistic:
\`\`\` q = |mean_i - mean_j| / sqrt(MS_within * (1/n_i + 1/n_j) / 2) \`\`\`
The q statistic is then compared to a critical value from the Studentized range distribution at the chosen alpha level (default 0.05). If q exceeds the critical value, that pair differs significantly.
Tukey HSD is appropriate when group sizes are equal or roughly equal. It controls the Type I error rate across all comparisons simultaneously.
Assumptions
One-Way ANOVA relies on three assumptions:
**1. Normality.** The data within each group should be approximately normally distributed. With larger samples (n > 30 per group), the Central Limit Theorem makes ANOVA fairly robust to moderate departures from normality.
**2. Homogeneity of variance.** The population variance should be equal across all groups (homoscedasticity). You can check this with Levene's test or by comparing the largest to smallest group standard deviation — a ratio below 2 is generally acceptable.
**3. Independence.** Observations must be independent within and between groups. This is ensured by study design, not statistical tests.
If your data strongly violate normality or have very unequal variances, consider the Welch ANOVA (which does not assume equal variances) or a non-parametric alternative such as the Kruskal-Wallis test.
How to Use
1. Enter a name for each group (optional — defaults to Group 1, Group 2, etc.) 2. Paste or type your numeric values into the data field for each group, separated by commas or line breaks 3. Click **Add Group** to add a third, fourth, or fifth group (up to 5 total) 4. Click **Run ANOVA** to compute the F-statistic, p-value, effect size, and ANOVA table 5. Review the **Tukey HSD** table (shown when you have 3 or more groups) to identify which specific pairs differ significantly
FAQs
Q: What is the minimum sample size for ANOVA? A: There is no hard minimum, but each group should have at least 2 observations to estimate within-group variance. In practice, groups with fewer than 5–10 observations have low statistical power, meaning real differences may not be detected. Aim for at least 10–20 observations per group for reliable results.
Q: Can I use ANOVA with only two groups? A: Yes, but a two-sample t-test is equivalent and more commonly used for two groups. The F-statistic from a two-group ANOVA equals the square of the t-statistic, and the p-values are identical.
Q: What does it mean when ANOVA is significant but no Tukey pairs are significant? A: This can happen with borderline p-values. The overall F-test is less conservative than pairwise post-hoc tests, so occasionally the omnibus test rejects the null hypothesis while no individual pair crosses the Tukey threshold. The practical difference is usually small.
Q: Do all groups need the same number of observations? A: No. One-Way ANOVA and Tukey HSD handle unequal group sizes (unbalanced designs) correctly. The formulas account for different n values in each group.
Q: What is eta-squared (η²)? A: Eta-squared is an effect size measure indicating the proportion of total variance explained by the group factor. A value of 0.10 means 10% of the variance in the outcome is attributable to group membership. Small = below 0.06, medium = 0.06–0.14, large = above 0.14.
Q: How is the p-value calculated without a table? A: The p-value comes from the F-distribution. This calculator uses the regularized incomplete beta function to compute the exact p-value numerically, the same method used by statistical software such as R and Python's SciPy.
Explore Similar Tools
Explore more tools like this one:
- Chi-Square Test Calculator — Perform chi-square goodness-of-fit and independence... - Quartile Calculator – IQR Calculator — Calculate quartiles (Q1, Q2, Q3) and interquartile range... - A/B Test Calculator — Determine the statistical significance of your... - P-Value Calculator — Determine statistical significance by calculating... - A/B Test Significance Calculator — Enter visitors and conversions for two variations to...