What does ANOVA tell you?

ANOVA tests whether the means of three or more groups are significantly different. A significant result means at least one group mean differs from the others, but doesn't tell you which. Use post-hoc tests (Tukey HSD) to identify specific differences.

What is the F-statistic?

The F-statistic is the ratio of between-group variance to within-group variance: F = MS_between / MS_within. A larger F-value indicates greater differences between groups relative to variation within groups. F is compared to F-distribution to get p-value.

When should I use ANOVA vs t-test?

Use t-test for comparing 2 groups. Use ANOVA for 3+ groups. Multiple t-tests inflate Type I error; ANOVA controls it. If you have multiple comparisons, use ANOVA first to test overall differences, then post-hoc tests for specific comparisons.

What assumptions does ANOVA make?

Independence of observations, approximate normality within each group, and homogeneity of variance (equal variances across groups). If violated: Welch's ANOVA (unequal variances), Kruskal-Wallis (non-normal), or transformations. Many analyses are robust to mild violations.

What's a post-hoc test?

A follow-up test after ANOVA to identify which specific groups differ. Tukey HSD is most popular — controls family-wise error rate. Bonferroni more conservative. Scheffé most conservative. Games-Howell handles unequal variances. Required when ANOVA is significant.

How do I calculate effect size for ANOVA?

Eta-squared (η²) = SS_between / SS_total. Proportion of total variance explained by group differences. η² = 0.06 small, 0.14 medium, >0.14 large. Always report with F and p. Cohen's f is another option for power analysis.

Can ANOVA be used for more than 3 groups?

Yes — this calculator handles 3 groups, but ANOVA works for any number of groups. For more groups: use statistical software (R, SPSS, Python). The principle is the same: compare group means relative to within-group variation.

One-Way ANOVA Calculator

Enter the mean, standard deviation, and sample size for 3 groups to perform a one-way ANOVA. Determines whether there are statistically significant differences between group means.

ANOVA (Analysis of Variance) is used to compare means across three or more groups. While t-tests work for two groups, comparing many groups with multiple t-tests inflates Type I error. ANOVA tests all groups simultaneously while controlling the overall error rate. The result is an F-statistic and p-value indicating whether at least one group mean differs from the others. ANOVA doesn't tell you which groups differ — that requires follow-up post-hoc tests.

This calculator performs one-way ANOVA for three groups given their means, standard deviations, and sample sizes. ANOVA partitions total variance into "between-group" (variance explained by group membership) and "within-group" (random noise). The F-statistic is the ratio of between to within variance. A large F suggests group differences are larger than random variation.

ANOVA assumes: independence (observations within and between groups), normality (data approximately normal within groups), and homogeneity of variance (similar variability across groups). When assumptions are violated, alternatives exist: Welch's ANOVA (unequal variances), Kruskal-Wallis (non-parametric), or transformations.

Common applications: comparing treatments in clinical trials, evaluating educational interventions across groups, testing manufacturing processes from different lines, A/B/C testing in marketing, and any multi-group comparison.

Inputs

Group 1 Mean

Group 1 Std Dev

Group 1 Size

Group 2 Mean

Group 2 Std Dev

Group 2 Size

Group 3 Mean

Group 3 Std Dev

Group 3 Size

Results

F-Statistic

2.5333

P-Value

0.015501

Decision

Significant (p < 0.05)

Grand Mean

27.6667

DF Between

DF Within

MS Between

63.3333

MS Within

25.0000

Last updated: May 29, 2026

Formula

**One-way ANOVA setup:** H₀: μ₁ = μ₂ = μ₃ (all means equal) H₁: at least one mean different **Total sum of squares:** SS_total = Σ(yᵢⱼ - ȳ_total)² **Between-group sum of squares:** SS_between = Σ nⱼ(ȳⱼ - ȳ_total)² **Within-group sum of squares:** SS_within = Σ Σ(yᵢⱼ - ȳⱼ)² **Degrees of freedom:** df_between = k - 1 (k = number of groups) df_within = N - k (N = total observations) df_total = N - 1 **Mean squares:** MS_between = SS_between / df_between MS_within = SS_within / df_within **F-statistic:** F = MS_between / MS_within **Critical F-value lookup:** Compare F to F-distribution with (df_between, df_within) and α. **Decision:** If F > F_critical (p < α): reject H₀ (significant difference). If F ≤ F_critical: fail to reject H₀. **Worked example: 3 groups (mean, SD, n)** Group 1: 25, 5, 10 Group 2: 30, 5, 10 Group 3: 28, 5, 10 Grand mean ≈ 27.67 SS_between = 10(25-27.67)² + 10(30-27.67)² + 10(28-27.67)² = 71.13 + 54.40 + 1.09 = 126.62 SS_within = 9(25) + 9(25) + 9(25) = 675 (using s² = 25 for each) df_between = 2; df_within = 27 MS_between = 63.31 MS_within = 25 F = 2.53 P ≈ 0.10 at α = 0.05; fail to reject (close but not significant). **ANOVA table:** | Source | SS | df | MS | F | p | |---|---|---|---|---|---| | Between | 126.62 | 2 | 63.31 | 2.53 | 0.10 | | Within | 675.00 | 27 | 25 | - | - | | Total | 801.62 | 29 | - | - | - | **Assumptions:** 1. **Independence**: observations don't affect each other. 2. **Normality**: data within each group approximately normal. 3. **Homogeneity of variance**: groups have similar variance (test with Levene's). 4. **Random sampling**: representative of populations. **Effect size:** - **Eta-squared (η²)**: SS_between / SS_total. Proportion of variance explained. - **Cohen's f**: standardized effect size. - Small: f = 0.10 - Medium: f = 0.25 - Large: f = 0.40 **Post-hoc tests:** When ANOVA is significant, identify which groups differ: | Test | Use | |---|---| | Tukey HSD | Pairwise comparisons; controls family-wise error | | Bonferroni | Conservative pairwise | | Scheffé | Multiple comparisons including contrasts | | Fisher's LSD | Less conservative | | Games-Howell | Unequal variances | **Welch's ANOVA:** When variances unequal, Welch's ANOVA is more robust. Most statistical software offers this option. **Kruskal-Wallis:** Non-parametric alternative when normality violated. Uses ranks instead of values. **Common applications:** - **Clinical trials**: compare drug responses across groups. - **Education research**: teaching methods comparison. - **Marketing**: A/B/C testing. - **Manufacturing**: process comparison. - **Agriculture**: crop yield by treatment. - **Psychology**: experimental conditions. **Sample size calculation:** Required n depends on: - Effect size (Cohen's f). - Number of groups (k). - Significance level (α). - Power (1 - β). Software (G*Power, R pwr) calculates required n. **Common errors:** - Multiple comparisons without correction. - Using ANOVA when assumptions violated. - Confusing significant F with which groups differ. - Forgetting effect size with statistical significance. - Using one-way when factorial needed. **Two-way ANOVA:** For two independent variables affecting outcome. Tests: - Main effects of each variable. - Interaction between variables. **Repeated measures ANOVA:** For comparing same subjects under different conditions. **Mixed ANOVA:** Combines between-subjects and within-subjects factors.

How to use this calculator

Enter mean, SD, and sample size for each of 3 groups.
Calculator returns F-statistic and p-value.
Compare p to significance level (α = 0.05 typical).
If significant: use post-hoc tests to identify differing groups.
Check assumptions: normality, equal variance, independence.
For more groups: use statistical software (R, SPSS, Python).

Worked examples

Drug effectiveness comparison

**Scenario:** Three drugs tested for blood pressure reduction. Means (mmHg): 12, 18, 15. SD: 4 each. n=20 per group. **Calculation:** Grand mean = 15. SS_between ≈ 720. F ≈ 5.0. p ≈ 0.009. **Result:** Significant differences exist (p < 0.05). Use Tukey HSD to identify: probably Drug 2 differs significantly from Drugs 1 and 3, while Drugs 1 and 3 may not differ significantly. Effect size moderate; clinically relevant.

Educational intervention

**Scenario:** Compare 3 teaching methods. Mean scores: 75, 78, 76. SD ≈ 8 each. n = 30 per group. **Calculation:** SS_between is small relative to SS_within. F ≈ 1.5. p > 0.05. **Result:** No significant differences detected. Sample size 30 per group has limited power to detect small differences. Larger study needed or effect is genuinely small.

Manufacturing process comparison

**Scenario:** Three production lines. Mean weight: 100.5, 100.2, 100.8 g. SD: 1.0 each. n=50 per line. **Calculation:** Grand mean = 100.5. F ≈ 8.0. p < 0.001. **Result:** Highly significant. Lines differ in mean weight. Investigate which lines need adjustment. Use post-hoc to identify specifically.

When to use this calculator

**Use one-way ANOVA for:**

- **Comparing 3+ group means** simultaneously. - **Single independent variable** with multiple levels. - **Continuous outcome variable**. - **Avoiding multiple t-test inflation** of Type I error.

**Use t-test instead when:**

- Comparing only 2 groups.

**Use two-way ANOVA when:**

- Two independent variables. - Want to test interaction.

**Use Welch's ANOVA when:**

- Unequal variances across groups.

**Use Kruskal-Wallis when:**

- Non-normal distributions.

**Post-hoc decisions:**

If overall ANOVA significant, use Tukey HSD (most popular) for pairwise comparisons. Bonferroni more conservative. Games-Howell handles unequal variances.

**Software:**

- **Excel**: Data Analysis ToolPak. - **R**: aov() function; TukeyHSD() for post-hoc. - **Python**: scipy.stats.f_oneway; statsmodels. - **SPSS**: Analyze → Compare Means → One-Way ANOVA.

**Common errors:**

- Comparing two groups with ANOVA (use t-test). - Multiple t-tests without correction. - Ignoring assumptions. - Confusing significant F with which groups differ. - Forgetting effect size.

**Effect size reporting:**

Report η² (proportion variance explained) or Cohen's f alongside p-value: - η² < 0.01: trivial - 0.01 - 0.06: small - 0.06 - 0.14: medium - > 0.14: large

**Power and sample size:**

Plan sample size for adequate power (typically 80%) before study. Software helps calculate based on expected effect.

**Reporting:**

Standard format: "A one-way ANOVA revealed a significant effect of [factor] on [outcome], F(df_between, df_within) = X, p = Y, η² = Z."

Then describe post-hoc findings.

**ANOVA vs t-tests:**

For k groups, doing all pairwise t-tests = k(k-1)/2 tests. - 3 groups: 3 tests; α inflated to ~14%. - 4 groups: 6 tests; α inflated to ~26%. - ANOVA tests all simultaneously at α = 0.05.

**Beyond one-way:**

- **Two-way ANOVA**: two factors (e.g., gender + treatment). - **MANOVA**: multiple dependent variables. - **ANCOVA**: covariates included as control. - **Repeated measures**: same subjects measured multiple times. - **Mixed models**: combines fixed and random effects.

**Practical guidelines:**

- **Always plot data** (box plots, scatter). - **Check assumptions** with diagnostic tests. - **Report effect size** with significance. - **Use post-hoc tests** to identify specific differences. - **Consider practical importance** beyond statistical significance.

Common mistakes to avoid

Using ANOVA for 2 groups (use t-test).
Skipping assumption checks (normality, equal variance).
Reporting significance without effect size.
Forgetting post-hoc tests when ANOVA significant.
Comparing means via multiple t-tests without correction.
Treating non-significant F as proof of no difference.
Misinterpreting F as which group differs.

Frequently Asked Questions

Sources & further reading

SponsoredShop Top Deals on AmazonSupport CalcMountain — browse top-rated products at no extra cost to you.