How to report a one-way ANOVA in APA style (7th ed.)
Three or more groups, one outcome. The sentence has more parts than a t-test sentence, and the part most often missing is the one that matters when the result is not significant. Here is what it needs, both patterns, a worked example, and six mistakes reviewers flag.
What the sentence must contain
The F statistic with both degrees of freedom — between groups, then within — the p-value, an effect size (η² or ω²), and the group means and standard deviations, in the text or in a table. If the overall test is significant, a post hoc test says which groups differ, with adjusted p-values.
The pattern when the result is significant
A one-way ANOVA showed that score differed by group, F(2, 57) = 4.21, p = .020, η² = .13. Tukey HSD comparisons indicated that Group A (M = 4.10, SD = 0.62) scored higher than Group C (M = 3.45, SD = 0.80), p = .015; the other pairs did not differ significantly.
Numbers in the pattern are placeholders.
Name the post hoc test. Report its adjusted p, not the unadjusted one. Give the means of the pairs that differ.
The pattern when the result is not significant
Score did not differ significantly by group, F(2, 57) = 1.42, p = .250, η² = .05.
Numbers in the pattern are placeholders.
Keep the effect size and the descriptives — a null result without them cannot be pooled or compared later. Write "did not differ significantly," not "there was no difference": the test did not find one, which is not the same thing.
A worked example
On a synthetic sample of 60 respondents in three departments — Research (M = 3.96, SD = 0.71), Sales (M = 3.51, SD = 0.80), Support (M = 3.70, SD = 0.79) — score did not differ significantly by department, F(2, 57) = 1.77, p = .180, η² = .06. Levene's test indicated equal variances.
This sentence was produced by the engine that computed the table; the same sample is in the six-step walkthrough.
Equal variances, and what to do when they are not
Report Levene’s test either way. If variances are unequal, use Welch’s ANOVA — F with a non-integer second df, F(2, 36.4) = 3.90 — and Games–Howell for the post hoc comparisons, and say so.
Six mistakes reviewers flag
- F without degrees of freedom — F = 4.21 tells the reader nothing about the design.
- "A trend toward significance" for p = .07 — report p = .07 and move on.
- Post hoc results with unadjusted p-values — say which correction, and use it.
- Effect size missing on a null result — that is exactly where it is needed.
- η² presented as partial η² or the reverse — they are equal only in a one-way design; name the one you report.
- "Explained 13% of the variance" without the interval — η² is a sample estimate; say so or give the CI.
Where the numbers come from
In Statory the sentence is generated from the same computation that produced the table, so the two cannot disagree. Every engine is checked line by line against established statistics software and the comparison is published — see the accuracy page.
The same engine runs inside ChatGPT and Claude when you paste a table there.