0
tutorials.scrollToStart
Statistical tests

When to Use ANOVA — and When to Use Something Else

6 min read

When to use ANOVA is straightforward once you know the rule: three or more groups, one metric dependent variable, and you want to compare means. ANOVA avoids the inflated error rate that comes from running multiple t-tests. This guide explains the difference between one-way, two-way, and repeated measures ANOVA, how to check its two key assumptions, and how to report η² in APA format.

Free sample chapter

Data Analysis From Survey to Results

Step-by-step guidance for choosing the right test, running it, and writing up APA results - in plain language, not theory. Get the free sample chapter when you join the waitlist.

Key takeaways

  • Use ANOVA when comparing means across three or more groups — never run multiple t-tests instead.
  • One-way ANOVA: one IV with 3+ levels. Two-way ANOVA: two IVs simultaneously. Repeated measures: same participants measured multiple times.
  • Check normality (Shapiro-Wilk) and homogeneity of variance (Levene’s test) before running ANOVA.
  • If Levene’s test is significant, use Welch’s F and Games-Howell post-hoc test.
  • Always run a post-hoc test (Tukey HSD or Games-Howell) to identify which specific groups differ.

What Is ANOVA? Comparing 3+ Groups Explained

ANOVA (Analysis of Variance) tests whether the means of three or more groups are significantly different from each other. Despite the name, it works by comparing variances — specifically, the variance between groups relative to the variance within groups (the F-ratio).

Typical thesis example: comparing stress levels across three study formats (in-person, hybrid, online). You have one independent variable (study format) with three levels, and one metric dependent variable (stress score). That is one-way ANOVA.

ANOVA tells you that at least one group differs — but not which ones. You need a post-hoc test for that.

One-Way vs. Two-Way vs. Repeated Measures ANOVA

Choose the ANOVA type based on your design:

  • One-way ANOVA: one independent variable with 3+ levels, different participants in each group.
  • Example: stress compared across three degree programmes.
  • Two-way ANOVA (factorial ANOVA): two independent variables simultaneously.
  • Example: stress compared by study format AND gender. Allows you to detect interaction effects.
  • Repeated measures ANOVA: the same participants measured at three or more time points.
  • Example: stress measured at start, mid-semester, and end of semester.
ANOVA typeIVsGroupsDesign
One-way1 IV3+ independent groupsBetween-subjects
Two-way2 IVsCells = Level(IV1) × Level(IV2)Between-subjects or mixed
Repeated measures1+ IVsSame participants, 3+ time pointsWithin-subjects

When to Use ANOVA vs. t-Test: The 3-Group Rule

The t-test handles exactly two groups. ANOVA handles three or more. The reason you cannot simply run multiple t-tests is statistical: each t-test carries a 5% Type I error risk. With three comparisons (A vs. B, A vs. C, B vs. C), the combined error rate rises to roughly 14%. ANOVA controls this by testing all groups simultaneously.

If normality fails: with n < 30 per group and Shapiro-Wilk p < .05, switch to Kruskal-Wallis (non-parametric ANOVA alternative).

If you have only two groups: use the t-test — a one-way ANOVA with two groups produces identical results.

Checking Assumptions: Normality and Levene’s Test in SPSS

ANOVA has two key assumptions:

1. Normality per group: Analyze → Descriptive Statistics → Explore → Dependent variable into Dependent List, grouping variable into Factor List → Plots → Normality plots with tests → OK. Check Shapiro-Wilk for each group (p > .05 = normality not violated).

2. Homogeneity of variance: run ANOVA via Analyze → Compare Means → One-Way ANOVA → Options → tick Homogeneity of variance test → OK. Levene’s p > .05 = variances are equal → use standard ANOVA. Levene’s p < .05 → use Welch’s F.

[TIP] With n > 30 per group, ANOVA is robust to mild normality violations due to the Central Limit Theorem. Check the Q-Q plot visually and proceed if deviations are minor.

Post-Hoc Tests: Tukey HSD Explained Simply

ANOVA’s F-test tells you that at least one group differs. A post-hoc test tells you which pairs differ.

Tukey HSD (Honest Significant Difference): use when Levene’s test is non-significant (equal variances). It controls the family-wise error rate across all comparisons.

Games-Howell: use when Levene’s test is significant (unequal variances). More conservative but reliable.

SPSS path: Analyze → Compare Means → One-Way ANOVA → Post Hoc → tick Tukey (or Games-Howell) → OK.

In the output, look at the Multiple Comparisons table: each pair shows a mean difference, standard error, and p-value. Pairs with p < .05 are significantly different.

Reporting η² as Effect Size in APA Format

  • Significant ANOVA:
  • "A one-way ANOVA revealed a significant effect of study format on stress scores, F(2, 118) = 7.43, p = .001, η² = .11." (η² = .11 is a medium effect; benchmarks: .01 small, .06 medium, .14 large.)
  • With post-hoc:
  • "Tukey HSD post-hoc comparisons indicated that students in the in-person format (M = 21.3, SD = 4.9) reported significantly lower stress than online students (M = 25.7, SD = 5.4, p = .002), but did not differ from hybrid students (M = 23.1, SD = 5.1, p = .082)."
  • Non-significant:
  • "No significant difference in stress was found across study formats, F(2, 118) = 1.24, p = .293, η² = .02."

[INFO] SPSS calculates eta-squared as part of the ANOVA output when you tick ‘Effect size’ in the Options dialog. Partial eta-squared and eta-squared are identical for one-way ANOVA.

Frequently asked questions

When should I use ANOVA instead of multiple t-tests?

Use ANOVA whenever you are comparing three or more groups. Running multiple t-tests instead inflates the Type I error rate — the probability of finding a false significant result increases with each additional comparison. ANOVA tests all groups simultaneously and controls this error rate.

What is the difference between one-way and two-way ANOVA?

One-way ANOVA tests the effect of one independent variable on a metric dependent variable across three or more groups. Two-way ANOVA tests two independent variables simultaneously and also detects interaction effects — whether the effect of one variable depends on the level of the other.

What do I do if Levene’s test is significant in ANOVA?

A significant Levene’s test (p < .05) means the variance assumption is violated. Switch to Welch’s F-test in SPSS (Options → Welch) and use Games-Howell for post-hoc comparisons instead of Tukey HSD.

How do I interpret the F-statistic in ANOVA output?

The F-statistic is the ratio of between-group variance to within-group variance. A large F means group differences are large relative to random variation within groups. Always read it together with the p-value and df. The ANOVA output shows F(df_between, df_within): for example F(2, 118) = 7.43, p = .001.

When should I use Kruskal-Wallis instead of ANOVA?

Use Kruskal-Wallis when the normality assumption fails (Shapiro-Wilk p < .05) and sample sizes are small (n < 30 per group). With larger samples, ANOVA is robust to normality violations. Kruskal-Wallis tests rank distributions rather than means and is the non-parametric equivalent of one-way ANOVA.

Free tool

Not sure which statistical test to use?

Answer 5 quick questions about your research design and get the right test - with an explanation of why - in under two minutes.

Statoria Team

Statistics educators & software developers

We build Statoria to help bachelor and master students get through their thesis data analysis without stress. Our guides are written by researchers with experience in social science statistics and student supervision.

Related guides