ANOVA tests whether group means differ beyond random variation. We emphasize assumptions, practical effect sizes, and clear follow‑up comparisons.
Designs
- One‑way ANOVA: one factor with k levels
- Two‑way ANOVA: two factors; check main effects and interaction
- For repeated measures use RM‑ANOVA or mixed models
Assumptions
- Normality of residuals (visuals and tests)
- Equal variances (Levene test)
- Independence of observations by design
Reporting
- ANOVA table with F, df, p‑value
- Effect sizes: eta‑squared or partial eta‑squared with CIs
- Post‑hoc: Tukey HSD (or Dunnett vs control) with multiple‑testing control
When Assumptions Fail
- Welch ANOVA for unequal variances
- Kruskal‑Wallis as a nonparametric alternative
- Transformations or robust estimators as needed