Reporting ANOVA results follows a specific format: you need the F-statistic, degrees of freedom, p-value, and an effect size, all written in a standardized way that lets other researchers quickly evaluate your findings. The exact formatting depends on whether you’re running a one-way, factorial, or repeated measures design, but the core components stay the same. Here’s how to get each piece right.
The Basic Format
Every ANOVA result in text follows this template:
F(dfbetween, dfwithin) = X.XX, p = .XXX, ηp2 = .XX
A real example looks like this: “An analysis of variance showed that the effect of noise was significant, F(3, 27) = 5.94, p = .007, ηp2 = .40.” That single line gives your reader everything they need: how many groups were compared, how large the effect was, and how confident you can be that it’s real. Before or alongside this line, you also report the descriptive statistics (means and standard deviations) for each group.
Getting the Numbers Right
APA 7th edition has specific rules about decimal places and leading zeros that trip people up constantly.
Round the F-statistic, means, and standard deviations to two decimal places. Report exact p-values to two or three decimal places (for example, p = .03 or p = .006). The one exception: when your p-value drops below .001, write p < .001 rather than reporting the exact number.
The leading-zero rule is simple once you know the logic. If a statistic can theoretically exceed 1, put a zero before the decimal (0.51, for a mean of 0.51). If a statistic can never exceed 1, drop the zero. P-values and effect sizes like η2 fall into this second category, so you write p = .007, not p = 0.007. The F-statistic itself always gets italicized, as does p.
Including Effect Size
Journals increasingly require an effect size alongside your F-test, and for good reason. A significant p-value tells you the group differences probably aren’t due to chance, but it says nothing about how large those differences are. Effect size fills that gap.
The most commonly reported effect size for ANOVA is partial eta squared (ηp2), largely because software like SPSS provides it automatically. Cohen’s benchmarks offer rough guidance for interpretation: .01 is considered small, .06 medium, and .14 large. Keep in mind that these benchmarks were originally developed for simple between-group comparisons. Using them to interpret effect sizes in designs with covariates or repeated measures stretches beyond their intended purpose, so treat them as loose reference points rather than hard cutoffs.
An alternative called omega squared (ω2) corrects for a slight upward bias in eta squared, making it a technically more accurate estimate. In practice, the difference between the two is usually small, and most researchers stick with partial eta squared or generalized eta squared (ηG2). Whichever you choose, name it explicitly so your reader knows which measure you’re using.
Writing Up a One-Way ANOVA
For a one-way ANOVA, you need three things in your results section: the descriptive statistics for each group, the omnibus F-test, and (if significant) the post-hoc comparisons. Here’s what that looks like in practice:
“Participants in the high-intensity group reported greater fatigue (M = 6.32, SD = 1.48) than those in the moderate-intensity (M = 4.51, SD = 1.22) or low-intensity groups (M = 3.10, SD = 1.35). A one-way ANOVA revealed a significant effect of exercise intensity on fatigue, F(2, 45) = 12.74, p < .001, ηp2 = .36.”
Notice the structure: describe the pattern in plain language first, then provide the statistical backup. Your reader should be able to understand the finding from the sentence itself, with the numbers serving as verification.
Reporting Post-Hoc Comparisons
A significant omnibus F-test tells you that at least one group differs from the others, but not which ones. If you have three or more groups, you need post-hoc tests to pin down the specific pairwise differences. Name the post-hoc method you used (Tukey, Bonferroni, Games-Howell, etc.) and report the results for each comparison.
“Tukey post-hoc comparisons indicated that the high-intensity group reported significantly more fatigue than the low-intensity group (p < .001) and the moderate-intensity group (p = .02). The moderate- and low-intensity groups did not differ significantly (p = .11).”
You can report post-hoc results using p-values alone, or include mean differences and confidence intervals for more detail. The key is naming which specific groups differed and which did not.
Reporting a Two-Way (Factorial) ANOVA
Factorial designs require you to report multiple F-tests: one for each main effect, plus one for the interaction. The standard order is to report the interaction first, because it determines how you interpret the main effects. If the interaction is significant, the main effects on their own can be misleading.
Here’s how that sequence looks:
“There was no significant interaction between message discrepancy and source credibility, F(1, 24) = 1.03, p = .32, ηp2 = .04. The main effect of message discrepancy was significant, F(1, 24) = 44.40, p < .001, ηp2 = .65, indicating that the mean change score was significantly greater for large-discrepancy messages (M = 4.78, SD = 1.99) than for small-discrepancy messages (M = 2.17, SD = 1.25).”
When the interaction is not significant, you simply state that and move on to the main effects. When it is significant, you’ll typically break down the interaction with simple effects tests and describe the pattern (for example, “the effect of X was stronger at level A of Y than at level B”).
What to Report When Results Are Not Significant
Non-significant results get the same statistical detail as significant ones. You still report the F-statistic, degrees of freedom, exact p-value, and effect size. The only difference is the framing language.
“A one-way ANOVA showed no significant effect of diet type on weight change, F(2, 57) = 1.18, p = .31, ηp2 = .04.”
Resist the temptation to write “the results were insignificant” (that implies unimportant) or to bury non-significant findings. Stating “no significant effect” or “the effect did not reach significance” is standard phrasing. Always include the exact p-value rather than just writing “n.s.” or “p > .05.”
Repeated Measures ANOVA and Sphericity Corrections
Repeated measures designs have an extra reporting requirement. The sphericity assumption (essentially, that the variability of differences between all pairs of conditions is equal) is frequently violated, and most software runs Mauchly’s test to check. When sphericity is violated, you apply a correction that adjusts the degrees of freedom downward, making the F-test more conservative.
Report it like this: “Mauchly’s test indicated that the assumption of sphericity was violated, χ2(2) = 8.41, p = .015, so degrees of freedom were corrected using Greenhouse-Geisser estimates (ε = .72). The results showed a significant effect of time on performance, F(1.44, 30.24) = 7.52, p = .005, ηp2 = .26.”
Notice the degrees of freedom are no longer whole numbers. That’s the correction at work, and it signals to your reader that you’ve accounted for the violation. If sphericity is not violated, you simply report the standard degrees of freedom and can note that the assumption was met.
Using Tables for Complex Results
When you have many groups or a factorial design with several effects to report, a table is often clearer than dense in-text statistics. APA recommends using standard table formats rather than inventing your own layout.
A typical ANOVA summary table includes columns for the source of variation (e.g., group, error), degrees of freedom, the F-statistic, the p-value, and the effect size. A separate descriptive statistics table lists the means and standard deviations for each group or condition. In published examples from APA, these tables show the degrees of freedom built into the F column header itself, like F(1, 294), with η2 in a final column.
Even when you use a table, your text should summarize the key findings in words. The table holds the full statistical detail; the text tells the story.
Common Formatting Mistakes
- Writing p = .000. Statistical software often outputs .000, but a p-value is never truly zero. Write p < .001 instead.
- Forgetting to italicize. F, p, M, SD, and N are all italicized in APA style. Greek symbols like η are not.
- Rounding p-values to one decimal. Report two or three decimal places. Writing p = .0 or p = .3 does not give your reader enough precision.
- Omitting effect size. Many instructors and reviewers now consider this mandatory. Include it even if your software doesn’t compute it automatically.
- Leaving a space inside parentheses for degrees of freedom. Write F(2, 45), not F( 2, 45 ). A comma and a single space separate the two df values.