ANOVA explained: when and how to use it

By Dr. Rafiq Muhammad, MD, PhD · Updated June 2026

ANOVA — analysis of variance — tests whether three or more group means differ more than chance would predict. Despite the name, it’s about comparing means; it just does so by partitioning variance. It returns an F-statistic and a p-value — and a significant result means “at least one group differs,” not which one.

Why not just run lots of t-tests?

Because every t-test carries a ~5% false-positive risk, and they pile up. Compare four groups pairwise and you need six t-tests — the chance of at least one false alarm climbs toward 25%. ANOVA tests everything in a single step at a controlled error rate. That’s the whole reason it exists.

One-way vs two-way

• One-way ANOVA — one factor (e.g. four teaching methods).
• Two-way ANOVA — two factors at once (e.g. method and gender), and it can test their interaction — whether the effect of one depends on the other.

You still need a post-hoc test

A significant ANOVA says the groups aren’t all equal but not which pairs differ. That’s what post-hoc tests (Tukey’s HSD, Bonferroni) are for — they find the specific differences while keeping the overall false-positive rate in check. Report the F-statistic, the p-value, an effect size (η² or partial η²), and the post-hoc comparisons.

Assumptions

Roughly normal residuals, independent observations, and similar variances across groups (homogeneity of variance). If variances differ badly, use Welch’s ANOVA; if normality is badly violated, the Kruskal–Wallis test is the non-parametric alternative.