Repeated Measures ANOVA Calculator

What Is Repeated Measures ANOVA?

Repeated measures ANOVA tests whether means differ across three or more conditions when the same subjects are measured in all conditions.

Question answered: "Do these conditions produce different outcomes for the same individuals?"

Examples:

• "Do patients' pain levels differ at baseline, 1 week, and 4 weeks after treatment?"
• "Do participants perform differently under easy, medium, and hard task conditions?"
• "Do reaction times change across morning, afternoon, and evening measurements?"

Why "Repeated Measures"?

Each subject provides multiple measurements:

This is within-subjects design, versus between-subjects (different people in each condition).

Advantages Over Between-Subjects

Controls Individual Differences

Each person serves as their own control. Variability due to individual differences is removed from error.

Between-subjects: High scorer in Condition A and low scorer in Condition B may obscure real effect.

Repeated measures: Compare each person to themselves across conditions.

Increased Statistical Power

With individual differences controlled, smaller true effects are detectable.

Same effect size requires fewer total subjects than between-subjects design.

Practical Efficiency

Fewer subjects needed. Valuable when participants are expensive, rare, or difficult to recruit.

The Statistical Model

Unlike one-way ANOVA, variance is partitioned into:

• Subject variance: How individuals differ (removed from error)
• Condition variance: Differences between conditions (our focus)
• Error variance: Subject × Condition interaction (residual)

This separation is why repeated measures has more power—subject variance doesn't contaminate error.

Calculating the F-Statistic

F = MS_Condition / MS_Error

Where n = subjects, k = conditions.

The Sphericity Assumption

Repeated measures ANOVA assumes sphericity: equal variances of differences between all pairs of conditions.

Why it matters: Violations inflate Type I error (false positives).

Testing Sphericity

Mauchly's test:

• Significant p < 0.05 indicates violation
• Non-significant suggests sphericity OK

Problem: Mauchly's test has low power with small samples when you most need it.

Correcting Violations

Apply correction factors to degrees of freedom:

Greenhouse-Geisser (ε_GG): Conservative. Recommended when sphericity is violated.

Huynh-Feldt (ε_HF): Less conservative. May be preferred when ε_GG > 0.75.

Lower-bound: Most conservative (ε = 1/(k-1)). Rarely used.

Corrected df: df_corrected = ε × df_original

Example Analysis

Scenario: Pain ratings (0-10) at three time points for 20 patients.

Mauchly's W = 0.82, p = .08 (sphericity OK)

Conclusion: Pain ratings differ significantly across time points.

Post-Hoc Comparisons

Significant F indicates conditions differ somewhere. Determine which:

Bonferroni-corrected paired t-tests: Compare each pair, adjusting α.

Polynomial contrasts: Test linear, quadratic, cubic trends (useful for time series).

Example: Significant linear trend means pain decreases steadily over time.

Effect Size: Partial Eta-Squared

η²p = SS_Condition / (SS_Condition + SS_Error)

Proportion of non-subject variance explained by condition.

In example: η²p = 98.4 / (98.4 + 119.8) = 0.45 (large effect)

Applications

Longitudinal Studies

Tracking changes over time in the same individuals.

Growth curves, treatment effects, developmental changes.

Within-Subject Experiments

Each participant experiences all experimental conditions.

Counterbalancing order prevents carryover effects.

Crossover Clinical Trials

Each patient receives all treatments (with washout periods).

More efficient than parallel designs.

Psychophysics

Testing perception across stimulus conditions within individuals.

Learning Studies

Measuring performance across practice sessions.

Design Considerations

Order Effects

Practice effects: Performance improves regardless of condition. Fatigue effects: Performance deteriorates over time. Carryover effects: One condition affects response to subsequent conditions.

Solutions:

• Counterbalance condition order across subjects
• Randomize order
• Include washout periods between conditions

Missing Data

Missing observations complicate repeated measures analysis.

Options: listwise deletion (lose power), mixed models (preserve data).

Sphericity in Depth

Technical definition: The variance of differences between conditions is constant for all pairs.

If Condition 1-2 difference has variance 4, Condition 1-3 difference should also have variance 4, as should Condition 2-3.

Compound symmetry (equal variances and covariances) implies sphericity but is stronger.

When violated: F-statistic is positively biased (too many false positives).

Mixed Designs

Combine between-subjects and within-subjects factors:

Example: Two groups (treatment vs. control) measured at three time points.

• Between-subjects factor: Group
• Within-subjects factor: Time
• Interaction: Group × Time (do groups change differently over time?)

The Group × Time interaction is often the key research question.

Assumptions

Independence

Observations from different subjects are independent.

Normality

Residuals approximately normal, OR adequate sample size.

Sphericity

Equal variances of pairwise differences.

Use corrections (Greenhouse-Geisser) when violated.

Common Mistakes

Ignoring Sphericity

Always check and report sphericity test; apply corrections when needed.

Treating as Between-Subjects

Using one-way ANOVA instead of repeated measures loses power and violates independence.

Ignoring Order Effects

Not counterbalancing in experiments allows confounds.

Over-Interpreting Without Post-Hocs

Significant F only means conditions differ somewhere.

Reporting Results

APA format (sphericity assumed):

"Repeated measures ANOVA revealed a significant effect of time on pain ratings, F(2, 38) = 15.6, p < .001, η²p = .45. Mauchly's test indicated sphericity was met, W = 0.82, p = .08."

With correction:

"Mauchly's test indicated sphericity violation, W = 0.65, p = .02. Greenhouse-Geisser corrected results showed a significant time effect, F(1.52, 28.9) = 12.3, p = .001."

Historical Context

Repeated measures ANOVA developed from Fisher's ANOVA framework but required addressing the sphericity issue.

Box (1954) and Geisser and Greenhouse (1958) developed corrections for sphericity violations that remain standard today.

The method became essential for psychological and biomedical research where within-subject designs are common and practical.

Related Calculations

• One-Way ANOVA — Between-subjects
• Two-Way ANOVA — Two factors
• Paired T-Test — Two conditions
• Friedman Test — Non-parametric alternative

Summary

Repeated measures ANOVA tests differences across conditions when the same subjects are measured repeatedly. By removing between-subject variance from error, it provides more power than between-subjects designs. The key assumption is sphericity—equal variances of pairwise differences. When violated (check with Mauchly's test), apply Greenhouse-Geisser or Huynh-Feldt corrections. Always conduct post-hoc comparisons to determine which conditions differ. Report corrected F-values when appropriate, effect sizes, and post-hoc results.