One-way ANOVA: testing group means
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
- The same patients appear at several visits. Use a repeated-measures model. Read: Choosing an approach for repeated measures
- Group spreads differ substantially. Consider Welch’s ANOVA, discussed in the tests-for-means overview. Read: Choosing a test for means
- You need baseline adjustment or other covariates. Use linear regression. Read: Linear regression
Before you start
What you need
- Raw values labeled by groupEach person belongs to one group and contributes one outcome.
- The clinical comparison and planned follow-up contrastsSpecify which differences matter and how several comparisons will be handled.
- The outcome and its unitsReport each group’s mean and spread, together with the relevant mean differences.
- Observations are independent within and between groups.How to check: Check the sampling and allocation units. Ordinary ANOVA does not account for shared wards.
- Classic ANOVA assumes a common population variance.How to check: Plot spread by group. Unequal sizes combined with unequal spreads deserve particular care.
- 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.
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.
Worked example
A worked example
- 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
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.
one-way-anova-tukey).analysis.one_way_anova, the id a plan cue records.Work it out
Related Learn pages
Related designs and methods
Evidence
Sources & further reading
- Non-normal data: Is ANOVA still a valid option? (opens in a new tab) · Psicothema · doi 10.7334/psicothema2016.383
- 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