Reporting
How to Report Mann-Whitney U Test Results in a Research Paper
Statistics for clinical researchers and surgical trainees
In short
Report five things: the U statistic, the exact p-value, the effect size r, the median with its IQR for each group, and the number of people in each group. Use the median and IQR, not the mean and standard deviation.
The Mann-Whitney U test is the most widely used non-parametric test in clinical and surgical research. It is also the one most often reported incorrectly, which makes it a frequent cause of revision requests from reviewers who know what to look for. The format is not complicated once you know the five pieces it needs.
The five things you must report
A complete Mann-Whitney U report has five parts. The first is U, the test statistic itself. The second is the exact p-value. The third is the effect size, written r, which describes how large the difference is. The fourth is the median and the interquartile range (the IQR, the middle 50% of the data) reported separately for each group. The fifth is n, the number of people in each group. Leaving any of these out will draw a comment from a careful reviewer.
The effect size formula
The effect size for this test is r = Z divided by the square root of N, where Z is a standardised version of the test statistic and N is the total number of people. This r is not the same as a correlation coefficient — it describes the size of the difference between two groups. The usual labels are r around 0.1 small, around 0.3 medium, and around 0.5 large; these thresholds come from Cohen.1 Report it next to U and the p-value every time. The Mann-Whitney U test itself is described in detail by Nahm.2
APA format
APA style: "Group A (Mdn = X, IQR = X–X) scored significantly higher [or lower] than Group B (Mdn = X, IQR = X–X), U = XXX, p = .0XX, r = .XX." For example: "Serum bilirubin was significantly higher in patients with hepatomegaly (Mdn = 2.55 mg/dL, IQR = 1.10–5.80) than in those without (Mdn = 1.00 mg/dL, IQR = 0.60–1.90), U = 17635.5, p < .001, r = .39." APA drops the zero before the decimal in the p-value and the effect size.
Vancouver format
Vancouver style: "Median (IQR): Group A X (X–X) versus Group B X (X–X); Mann-Whitney U = XXX, p = 0.0XX, effect size r = 0.XX." Vancouver keeps the leading zero and the result often appears as a footnote to a results table.
JAMA format
JAMA style italicises U and the capital P, keeps the zero before the decimal in U, and removes it in the p-value: "U = 17635.5, P < .001" (the effect size r = 0.39 may be appended). JAMA usually shows the median and IQR in square brackets.
| Group | n | Median | IQR |
|---|---|---|---|
| Hepatomegaly present | 160 | 2.55 mg/dL | 1.10–5.80 |
| Hepatomegaly absent | 152 | 1.00 mg/dL | 0.60–1.90 |
APAU = 17635.5, p < .001, r = .39
VancouverMann-Whitney U = 17635.5, p < 0.001, r = 0.39
JAMAU = 17635.5, P < .001
Example output. Figures are illustrative.
Example data: Vanderbilt University Department of Biostatistics public teaching datasets (hbiostat.org/data). Figures computed with scipy from real data.
Reporting it in a results table
| Variable | Hepatomegaly present (n=160) | Hepatomegaly absent (n=152) | U | p | r |
|---|---|---|---|---|---|
| Serum bilirubin (mg/dL) | 2.55 [1.10–5.80] | 1.00 [0.60–1.90] | 17635.5 | <.001 | .39 |
| Prothrombin time (s)† | 11.0 [10.5–12.0] | 10.6 [10.2–11.1] | — | — | — |
Values are shown as median [IQR], with the exact p-value and the effect size in the last columns. The bilirubin row is computed from the PBC dataset; † the prothrombin time row is illustrative, shown only to demonstrate the multi-row layout, and is not a computed result.
Common mistakes
Do not report the mean and standard deviation for data analysed this way — non-parametric results are summarised with the median and IQR. Do not leave out the effect size, which is the single most common reason these results get sent back. Do not write p = 0.000; write p < .001. Report U, not the intermediate z value. And avoid the phrase "approaching significance" for a result like p = .06 — it has no real meaning.
Try it yourself
Reproduce this result — in StatsPlease or SPSS
The result above comes from a public dataset, so you can compute U yourself in either tool and then practise formatting it for each journal style.
In StatsPlease
- Download the PBC dataset (see Data Sources) and save it as CSV.
- Upload it and choose serum bilirubin as the outcome and hepatomegaly as the group.
- Run. StatsPlease returns U, the exact p-value, r, and the median [IQR], pre-formatted for your target journal.
In SPSS
- Open the same CSV in SPSS.
- Go to Analyze ▸ Nonparametric Tests ▸ Independent Samples (Mann-Whitney U).
- Read U and the p-value; then add the median [IQR] and the effect size r by hand.
Compare: both should return U = 17635.5 and p < .001. SPSS gives you U and p; StatsPlease also computes r and writes the APA, Vancouver, and JAMA sentences for you.
You might also read
References
- Cohen J. Statistical Power Analysis for the Behavioral Sciences. 2nd ed. Hillsdale, NJ: Lawrence Erlbaum Associates; 1988.
- Nahm FS. Nonparametric statistical tests for the continuous data: the basic concept and the practical use. Korean Journal of Anesthesiology. 2016;69(1):8–14. https://doi.org/10.4097/kjae.2016.69.1.8
StatsPlease reports U, the exact p-value, the effect size, and median [IQR], formatted for your target journal.
Get Formatted Output →