One-Way ANOVA Calculator
Perform a complete One-Way Analysis of Variance (ANOVA) across 2 to 6 treatment groups. Computes the exact F-statistic, p-value, complete Sum of Squares ANOVA table, effect size (η², ω²), and interactive 95% confidence interval error bar charts.
Treatment Groups
Enter observations separated by commas or spaces.
ANOVA Summary Table
| Source of Variation | Sum of Squares (SS) | Degrees of Freedom (df) | Mean Square (MS) | F-Value | p-Value |
|---|---|---|---|---|---|
| Between Groups (Treatment) | 425.20 | 2 | 212.60 | 45.32 | < 0.0001 |
| Within Groups (Error) | 56.40 | 12 | 4.70 | - | - |
| Total | 481.60 | 14 | - | - | - |
Group Descriptive Statistics
| Group Name | Sample (n) | Mean (x̄) | Std Dev (s) | Std Error (SE) | 95% CI |
|---|
Group Means with 95% Confidence Intervals
Red dashed line = Grand MeanHow to Calculate One-Way ANOVA
To calculate One-Way ANOVA, partition total variation into between-group variance (MS_Between) and within-group variance (MS_Within). Divide MS_Between by MS_Within to determine the F-statistic, then evaluate the p-value against significance level alpha.
The Foundation of Analysis of Variance (One-Way ANOVA)
In experimental research and comparative data science, investigators frequently evaluate whether different treatments, environmental conditions, or drug dosages produce different average outcomes. While a Student's two-sample t-test compares exactly two groups, comparing three or more groups requires Analysis of Variance (ANOVA), introduced by statistician Sir Ronald Fisher in 1918.
If the null hypothesis is true (all group population means are identical), the sample group means only fluctuate due to random sampling noise, making the numerator and denominator approximately equal (F ≈ 1). If the treatment has a real effect, the between-group variation exceeds the within-group noise, driving F > 1 into the critical rejection region.
Reflects differences between the individual group sample means (x̄_i) and the overall grand mean (x̄). Captures both treatment effects and random noise.
Reflects variability among individual observations (x_ij) within the same group around their group mean (x̄_i). Represents pure experimental error and individual variation.
Mathematical Derivation & Sum of Squares Partitioning
The core algebraic property that empowers ANOVA is the Partitioning of Total Sum of Squares (SS_Total) into two orthogonal, mutually exclusive components:
Measures total variation of every individual observation from the grand mean:
Measures the weighted squared deviations of group sample means from the grand mean:
Measures the pooled deviations of individual data points from their respective group mean:
Hypothesis Formulation & Statistical Decision Rules
All treatment population means are identical. Any observed differences between group samples are purely due to random chance.
At least one population mean differs from the others. ANOVA is an omnibus test and does not indicate which specific pair differs.
- If p ≤ α (or F ≥ F_critical): Reject H₀. Conclude that there is statistically significant evidence of a difference between group means. Proceed to post-hoc tests.
- If p > α (or F < F_critical): Fail to reject H₀. There is insufficient evidence to conclude the means differ.
Quantifying Effect Size: Eta-Squared (η²) vs. Omega-Squared (ω²)
While p-values establish whether an effect is unlikely to be zero by chance, effect size metrics measure the magnitude and practical significance of the treatment:
Represents the proportion of total variance explained by the treatment groups. Similar to R² in regression analysis.
Corrects for sample bias, providing an accurate, conservative estimate of variance explained in the wider population.
Key Statistical Assumptions & Robustness Checks
Every sample unit must be randomly selected and independent. Paired or repeated measurements must use Repeated Measures ANOVA instead.
The distributions within each treatment group should be approximately normal. ANOVA is moderately robust to non-normality when group sample sizes are equal and n_i ≥ 20 (Central Limit Theorem).
All groups must share roughly equal population variances (σ₁² = σ₂² = ... = σₖ²). As a rule of thumb, if the ratio of highest group variance to lowest group variance is < 4:1 with equal sample sizes, standard ANOVA remains valid. Otherwise, use Welch's ANOVA.
Real-World Applications of One-Way ANOVA
Comparing blood pressure reductions across Placebo, 10mg, 25mg, and 50mg medication cohorts to identify minimum effective and optimal therapeutic dosages.
Evaluating crop yields (bushels per acre) across 4 different fertilizer formulations to determine which chemical blend significantly maximizes harvest volume.
Testing user conversion rates across three distinct landing page hero designs (Original, Video Hero, Minimalist) to optimize e-commerce funnel checkout speed.
Comparing tensile strength of aluminum components produced across 3 manufacturing plants to detect machine calibration drift and ensure quality uniformity.
Graded Step-by-Step Worked Example
A psychologist tests three teaching techniques (k = 3) on student exam scores (n = 5 per group, N = 15, alpha = 0.05):
- Group 1 (Lecture): 24, 28, 26, 30, 27 (Mean = 27.0, Sum = 135)
- Group 2 (Visual): 31, 35, 33, 36, 32 (Mean = 33.4, Sum = 167)
- Group 3 (Interactive): 38, 42, 39, 41, 40 (Mean = 40.0, Sum = 200)
SS_B = 5(-6.467)² + 5(-0.067)² + 5(6.533)² = 209.11 + 0.02 + 213.41 = 422.53
df_Between = 3 - 1 = 2 → MS_Between = 422.53 / 2 = 211.27
df_Within = 15 - 3 = 12 → MS_Within = 47.20 / 12 = 3.933
p-value = P(F(2, 12) > 53.71) < 0.0001 < 0.05 → Reject H₀. Teaching methods significantly affect exam performance.
Common Pitfalls in ANOVA Interpretation
A significant ANOVA only tells you that at least one pair differs. Group 1 and Group 2 might be identical, while Group 3 is vastly higher. Always inspect post-hoc tests.
Running pairwise t-tests without alpha correction inflates false positive error rates dramatically. ANOVA protects your statistical integrity.
If group sample sizes are unequal and variances differ by more than a factor of 4, standard ANOVA p-values can become inaccurate. Use Welch's ANOVA in such scenarios.
Lead Developer & Founder of Basic Math Tools. Specializes in browser-native computational algorithms and applied mathematics.
Mathematics & curriculum specialists. Audited against standard algebraic and arithmetic principles.