You ran your ANOVA, SPSS spat out a wall of tables, and now you need to turn that output into two or three sentences your professor will accept. This guide shows you exactly which numbers to pull, where they go, and how to format them in APA style (7th edition), with worked examples for one-way ANOVA, two-way ANOVA, and post-hoc tests. Copy the templates, swap in your own values, and you're done.
The Four Numbers Every ANOVA Report Needs
An APA-style ANOVA report always contains the same core pieces: the F statistic, two sets of degrees of freedom, the exact p value, and an effect size. Miss any one of these and most markers will flag it. The general template looks like this:
F(dfbetween, dfwithin) = value, p = value, η² = value
A real sentence built from that template reads: "There was a significant effect of teaching method on exam scores, F(2, 87) = 5.42, p = .006, η² = .11."
Each element follows specific formatting rules. The letter F is italicized, always. The two degrees of freedom sit inside parentheses, separated by a comma: the first is the between-groups df (number of groups minus 1), the second is the within-groups or error df (total sample size minus number of groups). The F value itself gets two decimal places. The p value gets two or three decimals with no zero before the decimal point, because p can never exceed 1. Write p = .006, never p = 0.006. When your software shows p as .000, report it as p < .001, since a probability is never exactly zero.
| Element | APA format | Common error |
|---|---|---|
| Test statistic | F(2, 87) = 5.42 | Missing df, or F not italicized |
| p value | p = .006 | p = 0.006 or p = .000 |
| Effect size | η² = .11 | Omitted entirely |
| Group means | M = 78.9, SD = 7.6 | Reported without SDs |
Before the statistics, your reader needs context. State which test you ran, what the independent variable was (including the number of levels), and what the dependent variable was. Descriptive statistics for each group, means and standard deviations, belong either in the text or in a table. If you have more than three groups, a table is cleaner. Our guide on writing a results section covers how to structure these paragraphs across a full chapter.
Every ANOVA sentence needs: F(df1, df2) = value, p = value, plus an effect size. Degrees of freedom are not optional, and neither is eta-squared.
Reporting a One-Way ANOVA: Worked Example
Suppose you compared final exam scores across three teaching methods: traditional lecture, flipped classroom, and problem-based learning, with 30 students in each condition. If you're unsure whether a one-way design was even the right choice for your data, the comparison in one-way vs two-way ANOVA walks through the decision. Your SPSS output shows an F of 5.42, a significance value of .006, and group means of 72.4, 78.9, and 76.2. Here is how the write-up should read:
"A one-way ANOVA was conducted to compare the effect of teaching method on final exam scores across three conditions: traditional lecture, flipped classroom, and problem-based learning. There was a significant effect of teaching method on exam scores, F(2, 87) = 5.42, p = .006, η² = .11. Students in the flipped classroom condition scored highest (M = 78.9, SD = 7.6), followed by the problem-based condition (M = 76.2, SD = 9.3) and the lecture condition (M = 72.4, SD = 8.1)."
Check the degrees of freedom against your design. Three groups gives dfbetween = 3 − 1 = 2. Ninety participants minus three groups gives dfwithin = 87. If your reported df don't match your sample size, examiners notice immediately, and it suggests you copied numbers without understanding them.
What if the result is not significant? Report it the same way, with the same completeness: "Teaching method had no significant effect on exam scores, F(2, 87) = 1.34, p = .267, η² = .03." Students sometimes drop the statistics for non-significant results, as if the numbers were embarrassing. Don't. A null result reported with full statistics is publishable science; a null result reported as "no difference was found" is an incomplete sentence.
If typing these strings by hand feels error-prone, the APA Results Formatter takes your F value, degrees of freedom, and p value and returns a correctly formatted sentence, with italics in the right places. It handles ANOVA along with t-tests, correlations, and chi-square.
Reporting Post-Hoc Tests
A significant omnibus F tells you that at least one group differs from at least one other. It does not tell you which. That's the job of post-hoc tests, and they need their own sentences in your results.
Tukey HSD
Tukey's Honestly Significant Difference test is the default choice when group sizes are equal and you want all pairwise comparisons. Report the pairs that differed, the direction, and the adjusted p value. Confidence intervals for the mean difference strengthen the report:
"Post-hoc comparisons using the Tukey HSD test indicated that the mean score for the flipped classroom condition (M = 78.9, SD = 7.6) was significantly higher than for the lecture condition (M = 72.4, SD = 8.1), p = .004, 95% CI [1.8, 11.2]. The problem-based condition did not differ significantly from either the flipped classroom condition (p = .412) or the lecture condition (p = .161)."
Notice that non-significant pairs are named too. Readers should not have to guess which comparisons you ran.
Bonferroni
The Bonferroni correction is stricter. It divides your alpha level by the number of comparisons, so with three pairwise tests each comparison is evaluated against .05 / 3 = .0167. When you use it, say so and explain the adjustment once:
"Pairwise comparisons with a Bonferroni correction (adjusted α = .0167) showed that flipped classroom scores exceeded lecture scores, p = .005, while the remaining comparisons were non-significant (both p > .10)."
One caution on software output: SPSS reports Bonferroni-adjusted p values that have already been multiplied by the number of comparisons, so you compare them against .05, not the divided alpha. State which convention you used. Mixing the two makes your thresholds impossible to verify.
Report post-hoc results for every pair you tested, significant or not, and name the correction method (Tukey, Bonferroni) explicitly. "Post-hoc tests were significant" on its own tells the reader nothing.
Reporting a Two-Way ANOVA and Interaction Effects
A two-way (factorial) ANOVA produces three F tests: one for each main effect and one for the interaction. All three get reported, whatever their significance. Say your study crossed teaching method (lecture, flipped, problem-based) with class size (small, large) using 120 students. The write-up:
"A 3 × 2 between-subjects ANOVA was conducted to examine the effects of teaching method and class size on exam scores. There was a significant main effect of teaching method, F(2, 114) = 6.88, p = .001, η²p = .108, and a significant main effect of class size, F(1, 114) = 4.21, p = .042, η²p = .036. These effects were qualified by a significant interaction between teaching method and class size, F(2, 114) = 3.67, p = .029, η²p = .060."
The phrase "qualified by" matters. When the interaction is significant, the main effects no longer tell a clean story, because the effect of one factor changes depending on the level of the other. Your next paragraph should unpack the interaction, usually with simple effects analyses:
"Simple effects analyses showed that the advantage of the flipped classroom over lecture was significant in small classes, F(1, 114) = 9.45, p = .003, but not in large classes, F(1, 114) = 0.71, p = .401."
Now the reader knows the practical meaning: flipped classrooms helped, but only when classes were small. A line graph of the cell means, with class size on the x-axis and separate lines for each method, makes the crossover visible at a glance and is worth including as a figure.
Report the interaction even when it fails to reach significance. "The interaction between teaching method and class size was not significant, F(2, 114) = 0.94, p = .394, η²p = .016" is a complete and necessary sentence, because it licenses you to interpret the main effects directly.
Effect Sizes: Eta-Squared and Partial Eta-Squared
The APA Publication Manual has required effect sizes since its 5th edition, and for good reason. With 90 participants, an F of 5.42 is worth attention; with 9,000 participants, a significant F can reflect a difference too small to matter in any classroom. The p value tells you whether an effect exists. The effect size tells you whether anyone should care.
For one-way ANOVA, report eta-squared (η²), which is the between-groups sum of squares divided by the total sum of squares. It represents the proportion of variance in the dependent variable explained by the grouping factor. An η² of .11 means teaching method accounted for 11% of the variance in exam scores. Cohen's rough benchmarks put .01 at small, .06 at medium, and .14 at large, though these cutoffs are conventions rather than laws, and you should interpret them against typical effects in your field.
For factorial designs, software typically reports partial eta-squared (η²p), which removes variance explained by the other factors from the denominator. Partial and classical eta-squared are identical in a one-way design but diverge in factorial ones, where partial values run larger. Whichever you report, name it correctly. Writing "η² = .108" when SPSS gave you a partial value misstates your result. The effect size guide explains the difference in depth, along with omega-squared, which corrects for the upward bias of eta-squared in small samples.
SPSS only prints partial eta-squared if you tick "Estimates of effect size" under Options, and some software (Jamovi's default output, older R functions) gives you sums of squares but no effect size at all. In that case you can compute η² yourself from the ANOVA table, or enter your sums of squares into the Effect Size Calculator, which returns eta-squared, partial eta-squared, and omega-squared together.
Place the effect size directly after the p value in the same statistical string. Burying it in a footnote or appendix defeats the purpose.
Common Mistakes in Student Write-Ups
After marking hundreds of results sections, the same errors keep appearing. Here are the ones that cost the most points.
Reporting p = .000
SPSS truncates small probabilities to three decimals, so a p value of .0000004 displays as .000. Copying that into your thesis is the single most recognizable sign that a student pasted output without thinking. Write p < .001.
Dropping the degrees of freedom
"The ANOVA was significant (F = 5.42, p < .05)" is incomplete twice over. Without df, the F value cannot be verified or converted to an effect size by a reader, and APA 7th edition asks for exact p values, not inequality thresholds, whenever p ≥ .001.
Calling p = .06 significant
If your alpha was .05, then .06 is not significant, and "marginally significant" or "approaching significance" are phrases most statisticians now reject, since p values do not approach anything. Report the result honestly as non-significant, give the effect size, and discuss the pattern in your discussion chapter if it seems meaningful.
Interpreting main effects past a significant interaction
In the two-way example above, concluding "flipped classrooms improve scores" ignores the interaction showing the benefit exists only in small classes. When the interaction is significant, lead with it and let simple effects carry the interpretation.
Running ANOVA on two groups, or t-tests on three
With exactly two groups, an independent-samples t-test and a one-way ANOVA give mathematically equivalent results (F = t²), but convention favors the t-test. With three or more groups, running multiple t-tests instead of ANOVA inflates your Type I error rate: three pairwise tests at α = .05 push the family-wise error rate toward 14%. The t-test vs ANOVA comparison covers when each test applies.
Inconsistent decimals and missing italics
APA convention is two decimal places for F, means, and standard deviations, and two or three for p values and effect sizes. The symbols F, p, M, and SD are italicized; the Greek η is not. Small details, but a marker skimming your results section registers them within seconds, and they signal whether the rest of your statistics deserve trust.
Forgetting assumption checks
Somewhere before the main result, mention that you checked homogeneity of variance (Levene's test) and normality. One sentence suffices: "Levene's test indicated equal variances across groups, F(2, 87) = 0.84, p = .435." If Levene's test is significant, report Welch's ANOVA instead and say so.
A Template to Adapt
Here is the full skeleton for a one-way design. Replace the bracketed parts with your own study details and numbers:
"A one-way ANOVA was conducted to compare the effect of [IV] on [DV] across [k] conditions: [list them]. Levene's test indicated equal variances, F(df1, df2) = X.XX, p = .XXX. There was a significant effect of [IV] on [DV], F(df1, df2) = X.XX, p = .XXX, η² = .XX. Post-hoc comparisons using Tukey's HSD indicated that [group A] (M = XX.X, SD = X.X) scored significantly higher than [group B] (M = XX.X, SD = X.X), p = .XXX. No other comparisons were significant (all p > .XX)."
Read your final version aloud once. If any sentence contains a statistic you could not explain to a classmate, go back to the output and work out where the number came from. That habit catches transcription errors, and it also happens to be excellent preparation for your defense, where "walk me through this F value" is a question examiners genuinely ask.