Use a repeated measures ANOVA when every participant is measured under all conditions and you have no between-groups factor; use a mixed ANOVA when you have both a within-subjects factor (the same people measured repeatedly) AND a between-subjects factor (different groups of people). The single question that decides it: do you have separate groups you want to compare, or just repeated measurements on one set of people? Getting this wrong is one of the most common reasons a thesis analysis gets sent back for revision.
Key Takeaways
- Repeated measures ANOVA analyses one group of participants measured across three or more time points or conditions — no separate groups involved.
- Mixed ANOVA combines a within-subjects factor (e.g. time) with a between-subjects factor (e.g. treatment vs control group) in one model.
- The deciding rule: if you have two or more independent groups that were each measured repeatedly, you need a mixed ANOVA.
- Both tests share the sphericity assumption; when it's violated, apply the Greenhouse-Geisser correction.
- The mixed ANOVA's interaction effect (group × time) is usually the result you actually care about — it tells you whether groups changed differently over time.
What is a repeated measures ANOVA?
A repeated measures ANOVA tests whether the mean of the same group of participants differs across three or more related conditions or time points. Because each person acts as their own control, it's more statistically powerful than comparing separate groups — it strips out individual differences.
Picture measuring anxiety in 30 patients before therapy, at week 6, and at week 12. Same 30 people, three measurements. There are no separate groups here — everyone follows the same path. That's a pure within-subjects design, and the repeated measures ANOVA asks one thing: did anxiety change significantly across those three time points?
If you only have two time points, you don't need ANOVA at all — a paired-samples t-test does the job.
What is a mixed ANOVA?
A mixed ANOVA tests whether two or more independent groups change differently across repeated measurements. It has at least one within-subjects factor and at least one between-subjects factor, which is why it's also called a "split-plot" or "between-within" ANOVA.
Now imagine the same anxiety study, but you randomly assign patients to a therapy group and a waitlist control group, and measure both at weeks 0, 6, and 12. You now have:
- A within-subjects factor: time (3 levels).
- A between-subjects factor: group (therapy vs control).
The mixed ANOVA gives you three effects: a main effect of time, a main effect of group, and — the headline result — the group × time interaction. That interaction answers the question a treatment study really asks: did the therapy group improve more than the control group over time?
Repeated measures ANOVA vs mixed ANOVA: key differences
| Feature | Repeated Measures ANOVA | Mixed ANOVA |
|---|---|---|
| Within-subjects factor | Yes (required) | Yes (required) |
| Between-subjects factor | No | Yes (required) |
| Typical question | Did scores change across conditions? | Did groups change differently over time? |
| Number of participant groups | One | Two or more |
| Key effect of interest | Main effect of the within factor | Group × time interaction |
| Example | Anxiety at 3 time points in one cohort | Anxiety over time in therapy vs control |
| Sphericity assumption | Applies | Applies (to within factor) |
The simplest way to remember it: a mixed ANOVA is a repeated measures ANOVA with a grouping variable added. If you have groups to compare, go mixed.
When should I use a mixed ANOVA instead of repeated measures?
Choose a mixed ANOVA whenever your design contains both types of factor at once. Concretely, use it when:
- You have experimental and control groups each tested at multiple points.
- You're comparing men vs women (or any grouping) across several conditions.
- Your core hypothesis is about a differential change — one group responding to an intervention while another doesn't.
Stick with a plain repeated measures ANOVA when every participant experiences every condition and there's no meaningful group to split them into. If you're unsure whether your grouping variable belongs in the model, StatRyx flags it automatically when you load your data and picks the correct ANOVA for you — no manual decision tree required.
A worked example with real numbers
Let's run the mixed design. In a study of 60 participants, 30 received cognitive behavioural therapy and 30 sat on a waitlist. Anxiety (GAD-7 scores) was measured at weeks 0, 6, and 12.
The mixed ANOVA returns:
- Main effect of time: F(2, 116) = 18.42, p < .001, η²ₚ = .24 — anxiety changed significantly across the three time points overall.
- Main effect of group: F(1, 58) = 3.11, p = .083, η²ₚ = .05 — averaged across time, the two groups didn't differ significantly.
- Group × time interaction: F(2, 116) = 9.76, p < .001, η²ₚ = .14 — the groups changed differently over time.
Here's what those numbers mean in plain English. The significant interaction is the finding that matters: it tells you the therapy group's anxiety dropped while the control group's stayed flat. The non-significant main effect of group is expected — at baseline both groups were similar, so their overall averages look alike. In a treatment study, the interaction is the result you report first and interpret hardest.
The partial eta squared (η²ₚ = .14) tells you the interaction explains roughly 14% of the variance after accounting for other factors — a medium-to-large effect by Cohen's benchmarks (.01 small, .06 medium, .14 large).
Don't forget sphericity
Both tests assume sphericity — that the variances of the differences between conditions are equal. Mauchly's test checks it. If it's significant (p < .05), apply the Greenhouse-Geisser correction, which adjusts the degrees of freedom downward to keep your p-value honest. This is one of the most-missed steps in student analyses, and it's applied automatically in StatRyx.
How do I report each one in APA 7?
For a repeated measures ANOVA:
A one-way repeated measures ANOVA showed a significant effect of time on anxiety, F(2, 58) = 12.30, p < .001, η²ₚ = .30.
For a mixed ANOVA, report all three effects, leading with the interaction:
A 2 × 3 mixed ANOVA revealed a significant group × time interaction, F(2, 116) = 9.76, p < .001, η²ₚ = .14. The main effect of time was significant, F(2, 116) = 18.42, p < .001, η²ₚ = .24, whereas the main effect of group was not, F(1, 58) = 3.11, p = .083, η²ₚ = .05.
Note the APA conventions: F and p italicised, no leading zero on p values, and effect sizes reported for every effect. If you're deciding which follow-up test to run after a significant interaction, see our guide on simple main effects and post-hoc comparisons.
Stop calculating this by hand — run it free in StatRyx → Try StatRyx
Frequently Asked Questions
Is a mixed ANOVA the same as a two-way repeated measures ANOVA?
No. A two-way repeated measures ANOV