Reporting Repeated Measures ANOVA in SPSS
Adjusted df and Follow-Up Results
Reporting repeated measures ANOVA in SPSS turns on which row of the within-subjects table is quoted. The F carries adjusted degrees of freedom when the epsilon adjustment applies, followed by partial eta squared and the follow-up comparisons.
IN SHORT
What a Within-Subjects Report Must Carry
A repeated measures report states whether the mean outcome changed across occasions or conditions in the same participants. It gives an F ratio with the degrees of freedom actually used, an effect size, and follow-up comparisons that locate the change, together with the mean at each occasion. For reporting repeated measures ANOVA in SPSS, the values come from Analyze, General Linear Model, Repeated Measures: the sphericity test, the Tests of Within-Subjects Effects table, the pairwise comparisons printed under estimated marginal means and the descriptive means. Labels differ between releases. OnlineSPSS is a specialist statistical consultancy for academic, clinical and professional research, and reconciling a within-subjects write-up with its saved output is work for its in-house PhD-level statisticians.
In This Guide
ANALYSIS SUPPORT
Analysis Support for This Method
Which model suits the repeated measurements decides how the result is written. An ordinal outcome, or one whose differences across occasions are far from normal, points to ranks instead.
Repeated Measures ANOVA
The analysis of one outcome across occasions in the same participants, run for your design, with the follow-up comparisons and the wording planned together.
Friedman Test
The rank-based analysis of three or more related measurements, for ordinal outcomes or when the model assumptions cannot be defended.
BEFORE YOU REPORT IT
What Fixes the Wording of the F
The way the F is written depends on four facts about the output, and each can be read before the paragraph is started.
One Row Per Adjustment The within-subjects table prints one row for each adjustment. The F is identical in every row, while the df and the p-value change, so one row is chosen and all its values are taken together.
Adjusted df As Printed The adjusted df are the unadjusted df multiplied by epsilon, which usually leaves decimals. They are reported as printed, with the epsilon value stated in the text.
Planned or All Pairs Contrasts specified before the data were seen are reported as planned. Pairwise comparisons across all occasions need an adjustment for multiplicity, and the type is named.
Complete Cases Only A participant missing any one occasion is dropped from the whole analysis. The N and the error df reflect that, and may fall below the number recruited.
IN SPSS
Taking Values From the Within-Subjects Tables
What the sphericity output means and how the row is chosen are explained in GLM Repeated Measures in SPSS: Output Guide, and the steps below take their values from the tables IBM SPSS prints.
Take the Epsilon Estimate
From Mauchly’s Test of Sphericity, take Mauchly’s W with its Sig., and the epsilon estimates printed in the same table.
Tip: Mauchly’s test has low power in small samples and flags trivial departures in large ones, so the epsilon value and the stated rule decide the row, not the test alone.
Take the F, Adjusted df and p
From Tests of Within-Subjects Effects, take F, df and Sig. from the adjusted row for the factor, and the df from the matching Error row.
Watch: The effect and Error rows each have their own df. Write F(df effect, df error) with both taken from the same adjustment row, never one value from each of two rows.
Take Partial Eta Squared
Where estimates of effect size were requested under Options, take Partial Eta Squared from the same row. It is not always the last column printed.
Watch: SPSS prints no confidence interval for it. Say so in the text, give the value as a point estimate, and call it partial eta squared, never eta squared.
Take the Follow-Up Results
Take pairwise Mean Difference (I-J) with its confidence interval and adjusted Sig. from Pairwise Comparisons, or the Tests of Within-Subjects Contrasts for planned contrasts.
Check: The default in the dialog applies no adjustment to pairwise comparisons. The footnote under the table names the one used, and the paragraph repeats that name.
Take the Means by Occasion
From Descriptive Statistics, take the N, mean and standard deviation at each occasion, or the estimated marginal means.
Tip: Take the N from this table as well. A participant missing one occasion leaves the whole analysis, so the N may be lower than the sample recruited.
Trace Each Figure to One Row
Write the F sentence, then the follow-up sentence, then fill the table, and set each against the saved output.
Check: Does F(df1, df2) show decimals once an adjusted row is quoted? Whole numbers beside an adjusted F usually mean that the unadjusted row was copied.
READING THE OUTPUT
How the Within-Subjects Tables Map to the Paragraph
The tables are listed in the order the paragraph uses them, from the epsilon estimate through the F to the pairwise rows and the means.
| Output value | Where SPSS prints it | Where it goes and what to check |
|---|---|---|
| Mauchly’s W and Sig. | Mauchly’s Test of Sphericity | A methods sentence on the check made, and never the only reason for choosing a row |
| Epsilon (Greenhouse-Geisser, Huynh-Feldt, Lower-bound) | Same table | The epsilon value stated beside the adjusted F, so the size of the adjustment is visible |
| F | Tests of Within-Subjects Effects | The same F in every row; it goes in the sentence once, with the df of the row chosen |
| df and Error df | Adjusted row and its Error row | F(df, error df), both taken from the same adjustment row, with their decimals kept |
| Sig. | Same row | The p-value for that row, not for another, with a printed .000 given as p < .001 |
| Partial Eta Squared | Same row | The effect size, labeled in full, with a note that the output gives no interval for it |
| Mean Difference (I-J), 95% CI, Sig. | Pairwise Comparisons | The follow-up result with its interval, and the adjustment named in the table footnote |
| N, Mean, Std. Deviation | Descriptive Statistics | The means-by-occasion table, with the N of complete cases beside the means |
REPORTING
The Adjusted F, Its Follow-Ups and a Means-by-Occasion Table
The sentence leads with the adjusted F, and the table lets the reader see each occasion's mean beside it.
In repeated measures ANOVA APA format, the F is written with both degrees of freedom, and the sphericity result goes into the sentence only through the epsilon adjustment. Reporting adjusted degrees of freedom means giving the decimals exactly as printed. The frame below is for one within-subjects factor.
A repeated measures ANOVA showed that [outcome] [changed / did not change] across [occasions], F(___, ___) = ___, p = ___, partial η² = ___, with a [Greenhouse-Geisser / Huynh-Feldt] epsilon adjustment to the degrees of freedom (ε = ___; row chosen by [rule stated in advance]). Pairwise comparisons with a [adjustment name] adjustment showed that [Time 1] differed from [Time 2] by a mean of ___, 95% CI [___, ___], adjusted p = ___. Means at each occasion are given in Table 1.
| Table 1. [Outcome] at Each Occasion, With Means and Standard Deviations | ||||
|---|---|---|---|---|
| Occasion | n | M | SD | Mean difference from [Time 1] (95% CI) |
| [Time 1] | ___ | ___ | ___ | reference |
| [Time 2] | ___ | ___ | ___ | ___ [___, ___] |
| [Time 3] | ___ | ___ | ___ | ___ [___, ___] |
| Note. F(___, ___) = ___, p = ___, partial η² = ___, with the epsilon adjustment named in the text (ε = ___). Pairwise p-values are adjusted by [method]. n is the number of complete cases. | ||||
Partial eta squared is the share of variability attributable to the factor once other effects are set aside, and the paper labels it as such. SPSS prints no interval for it, so an interval needs another program. Generalized eta squared, which is more comparable across designs, is not printed either. Benchmarks for small, medium and large effects are conventions, and the size is read against the measurement scale and the field. Each follow-up carries its own effect: the mean difference with its interval is the quantity readers use for a pair.
COMMON MISTAKES
How a Repeated Measures Paragraph Drifts From Its Output
Most of these appear because the within-subjects table has several rows that look alike, and each is found by tracing one value back to one row.
Whole-Number df Reporting the adjusted F with the unadjusted df makes the p-value belong to a different row than the F suggests. The decimals are kept as printed.
Values From Two Rows Taking df from one adjustment row and the p-value from another produces a result no row of the table contains. All values come from the same row.
Mauchly Alone Decides Choosing the row only because Mauchly's p crossed .05 ignores its power problems. The epsilon value and a rule stated in advance decide the row.
Adjustment Left Silent Follow-up comparisons with no named adjustment cannot be reproduced, and the default in the dialog applies none. The footnote of the table gives the name.
Partial Eta Squared Mislabeled Calling partial eta squared eta squared, or comparing it across designs with different error terms, misstates its meaning. The label is written in full.
Complete Cases Not Mentioned A report that gives the recruited sample while the analysis used fewer participants hides the loss. The N of complete cases belongs in the table.
Which row did the F come from? Independent Statistical Review follows each value in your paragraph back to a single row of the output and to the table that holds it.
FREQUENTLY ASKED QUESTIONS
Epsilon, the Multivariate Table and Partial Eta Squared
Decimal df, a multivariate block above the univariate rows and an effect size printed without an interval: each answer starts from what the output shows.
Why does my F have decimals in its degrees of freedom?
The epsilon adjustment multiplies both degrees of freedom by epsilon, which usually leaves decimals. The F value stays the same, while the df and the p-value change. Report the df as printed, name the estimate used exactly as its row is labeled in the output and give its epsilon value so the reader can see how large the adjustment was.
Do I report the Multivariate Tests table that SPSS prints first?
Only if the multivariate approach was the planned analysis. That table sits above the univariate rows and reports Pillai’s trace, Wilks’ lambda and related statistics. Name the one approach planned in advance, the multivariate test or the univariate F with an epsilon adjustment, and report that one, not whichever table gives the smaller p-value.
How do I report partial eta squared without an interval?
Give the point estimate, name it in full and note that SPSS prints no interval for it. If a journal expects an interval, it is calculated in another program from the F and the df. Avoid comparing the value with effect sizes from designs that have different error terms.
How is a repeated measures write-up reviewed against the output?
An in-house PhD-level statistician at OnlineSPSS follows the F, the df, the epsilon value, the effect size and each pairwise result back to the saved tables, then confirms the wording. It belongs to Results Reporting & Interpretation, while the reading of the change over time is argued in the paper’s own discussion.
How do you protect my research data and confidentiality?
Your data and project details are handled confidentially, and project information is not published. Researcher names and institutions are not used publicly without permission. An NDA is available where required and can be arranged before sensitive materials are shared. Read more about confidentiality and data security.
What happens when I request an assessment, and how long does it take?
You share your research question, design and relevant materials, such as the dataset or the output you have, and a statistically trained Research Coordinator reviews the request and responds the same day. The analysis, deliverables, timeline and one fixed project fee are then confirmed before any work begins, so the timeline follows from the agreed scope.
Still have a question? Find more answers in our full FAQ.
NEXT STEP
Does Your F Carry
the Adjusted df?
The adjusted F, its epsilon estimate and each follow-up row have to come from one saved output file, with the same complete cases behind every value in the paragraph.
No Commitment Required
Strictly Confidential
Fixed Project Fee
In-House PhD-Level Statisticians
Same-Day Response
Serving Researchers Since 2014