How to Report Statistical Analysis Results

Reporting statistics well isn’t about drowning the reader in numbers — it’s about presenting the few that answer the question, and saying plainly what they mean.

Once the analysis is done, you have to communicate it — and this is where many otherwise-sound projects lose marks. Good statistical reporting presents results accurately and clearly enough that a reader can follow what you found and judge whether your conclusions are justified. Precision and restraint matter more than volume; a results section is not a place to show every calculation you ran.

Describe your data first

Before any test result, orient the reader with descriptive statistics — sample size, and measures of central tendency and spread (mean and standard deviation for continuous data, counts and percentages for categorical data). This paints the picture your inferential tests then interpret, and it lets the reader sanity-check everything that follows. Mention any missing data or exclusions here too, so the numbers add up.

Report the test, not just the verdict

When you report an inferential result, name the test, give the key statistic, the degrees of freedom, and the p-value — and, increasingly expected, an effect size. “There was a significant difference in scores between groups” is not enough; report the actual numbers so the result is verifiable and its magnitude is clear. Follow the conventions of your field’s style guide (in psychology and many social sciences, that’s APA), including how to format statistics, italics, and decimal places.

Use tables and figures deliberately

Tables and figures should carry information that would be clumsy in prose — not repeat it. Put detailed numbers in a well-labelled table; use a figure when a pattern is easier to see than to read, such as a trend over time or a difference between groups. Give each a number and a caption, refer to it in the text, and make sure it stands alone and reads clearly in greyscale. Don’t present the same numbers three times — in a table, a figure, and the text.

Separate results from interpretation

In the results section, report what you found without explaining why. Save the meaning for the discussion, where you interpret the numbers in light of your research question and the existing literature. Keeping the two apart stops you from sliding into speculation before the reader has seen the evidence, and it makes the report easier to follow, because each section has one clear job.

Interpret significance honestly

Statistical significance is not the same as importance. A p-value below your threshold tells you a result is unlikely to be chance; it doesn’t tell you the effect is large or meaningful — that’s what effect sizes are for. Equally, a non-significant result is a genuine finding, not a failure to be hidden or quietly dropped. And correlation is not causation: report what your design can actually support, and no more. Overclaiming is one of the fastest ways to lose a reader’s trust and a marker’s confidence.

Write for a reader, not a calculator

Round sensibly rather than reporting six decimal places, define your symbols and abbreviations, and lead each result with the plain-language point before the numbers that back it — “scores were higher in the treatment group” and then the statistics. The aim is a reader who finishes your results section knowing exactly what you found and how confident they should be, not one buried under figures they have to decode on your behalf.

Report enough for a reader to judge the result

A statistical result is a claim about evidence, and the reader needs enough information to assess it. That generally means the test used, the test statistic with its degrees of freedom, the p-value, an effect size, and the descriptive statistics the test was run on. Omitting the effect size is the most common gap and the most consequential, because significance alone says nothing about magnitude.

Give descriptive statistics before inferential ones. A reader who does not know the group means cannot interpret a difference between them.

Distinguish significance from importance

A significant result means the data would be unlikely under the null hypothesis. It does not mean the effect is large, or that it matters. With a large enough sample, trivial differences reach significance; with a small one, important effects may not.

This is why effect sizes and confidence intervals carry the interpretive weight. Reporting that a difference is significant and then discussing its practical importance without reference to its size is a very common error, and one markers look for specifically.

Use the conventional format for your field

Most disciplines follow APA conventions closely, and the details are assessed. Statistical symbols are italicised, exact p-values are reported rather than thresholds where possible, and leading zeros are omitted from values that cannot exceed one.

• t-test: t(48) = 2.31, p = .025, d = 0.66
• ANOVA: F(2, 87) = 4.12, p = .020, η² = .09
• Correlation: r(98) = .34, p < .001
• Chi-square: χ²(3, N = 210) = 9.44, p = .024

Report p as “< .001” rather than “.000”, which is what software prints and is never literally true.

Choose tables and figures for the right job

Tables suit precise values and many comparisons; figures suit patterns, trends and interactions. Do not present the same data as both. Every table and figure needs a number, a caption that makes it interpretable without the surrounding text, and a mention in the text that says what it shows rather than merely pointing at it.

Where results will be printed in greyscale, check that a figure remains readable without colour — a distinction carried only by red and green disappears entirely.

Report what did not work

Non-significant results belong in the report. Omitting them distorts the picture and, in research contexts, is a recognised form of misconduct. Report them in the same format as significant ones, and resist describing a non-significant trend as “approaching significance” — a p-value of .07 either meets your stated threshold or it does not.

The same applies to assumption checks. Where a test assumed normality or homogeneity of variance, say whether you checked and what you found, and state which alternative you used if the assumption failed.

Keep interpretation for the discussion

Results sections state what was found; discussion sections say what it means. Sentences beginning “this suggests that” generally belong in the discussion. Keeping the boundary clean makes both sections easier to write and easier to mark, and it protects you from the frequent error of interpreting a result before the reader has seen it.