Ask Mallard
Analysis method

One-way ANOVA: testing group means

4 min read

One-way ANOVA tests whether independent groups share the same population mean; it does not identify the groups that differ.

Fit

When to use it, and when not to

Use it when

  • One categorical factor defines three or more independent groups with a continuous outcome.
  • You need an overall test of equal means, with any planned contrasts stated in advance.

Not when

Before you start

What you need

Decide these first
  1. Raw values labeled by groupEach person belongs to one group and contributes one outcome.
  2. The clinical comparison and planned follow-up contrastsSpecify which differences matter and how several comparisons will be handled.
  3. The outcome and its unitsReport each group’s mean and spread, together with the relevant mean differences.
It rests on these
  1. Observations are independent within and between groups.How to check: Check the sampling and allocation units. Ordinary ANOVA does not account for shared wards.
  2. Classic ANOVA assumes a common population variance.How to check: Plot spread by group. Unequal sizes combined with unequal spreads deserve particular care.
  3. Small-sample F inference assumes roughly normal errors within groups.How to check: Inspect deviations from group means for skew and extreme values.

Plain language

How it works

ANOVA means analysis of variance. The method compares variation between group means with variation within groups. Their ratio is the F statistic. A large ratio challenges the null that all population means are equal.

What the sources show

Blanca and colleagues studied normality and unequal variances in separate simulations. Robustness to non-normality does not remove the equal-variance assumption. The calculator runs classic ANOVA, not the unequal-variance Welch variant.

The method at a glance
One-way ANOVA: Independent outcome groups. Then Between and within group variation. Then Overall F test of equal means.One-way ANOVAIndependent outcome groupsBetween and within groupvariationOverall F test of equalmeans
Conceptual diagram. It shows the structure of the idea, not data from any study.

Worked example

A worked example

Invented for illustration
Setting
3 independent groups each provide 3 outcome measurements.
Question
Are all three population means equal?
Approach
Group 1: 1, 2, 3. Group 2: 2, 3, 4. Group 3: 4, 5, 6.
Result
The group means are 2, 3 and 5. The overall F = 7.00, df = 2 and 6, p = 0.027.
What it means
The overall test gives evidence against equal means. It does not identify a specific pair; report planned contrasts with suitable intervals.

In practice

Recent studies PubMed search · not vetted

Searching PubMed…

Watch out

Common mistakes

The one to remember

A significant overall F test does not identify which means differ. Classic one-way ANOVA assumes equal population variances; plan any follow-up comparisons and multiplicity control.

  • Running every pairwise t-test without an error-control plan.Instead: Prespecify the comparisons and their multiplicity adjustment.
  • Reporting only the F statistic.Instead: Give group means, sample sizes and spreads, plus the contrasts that answer the clinical question.

Inside Mallard

How Mallard uses this

Mallard links this article to its classic one-way ANOVA calculator. The tool reports an overall F test. The method registry links this method to its published analysis article and code.

Code example in Mallard: One-way ANOVA with Tukey pairwise intervals (one-way-anova-tukey).
Sample size: Mallard sizes the main comparison with anova k groups. It does not size adjustment or any secondary analysis.
Calculators: One-way ANOVA (§3.1).
Atlas node analysis.one_way_anova, the id a plan cue records.
Code for this method
Loading…

Work it out

Related Learn pages

Related designs and methods

Evidence

Sources & further reading

  1. Non-normal data: Is ANOVA still a valid option? (opens in a new tab) · Psicothema · doi 10.7334/psicothema2016.383
  2. Effect of variance ratio on ANOVA robustness: Might 1.5 be the limit? (opens in a new tab) · Behav Res Methods · doi 10.3758/s13428-017-0918-2
Prepared with AI assistance. Source records verified by PMID, title, authors, year and DOI on 2026-11-06. No independent professional approval is claimed. The two primary simulation studies separately evaluate non-normality and unequal variances. Their findings apply to the conditions studied, not every possible distribution.
Last source review: 2026-11-06
AI-generated educational content
Mallard uses reputable methodological and reporting sources, but this material may contain errors or omit important context. Verify important methodological decisions with the cited sources and, when appropriate, a qualified research methodologist or statistician.