Statistical Analysis Paragraph Generator: Write the Statistics Part of Your Methods (Free)
Choose your study design, the tests you ran, how you checked assumptions, your post-hoc tests and corrections, the software and version, and how you report the data. You get a complete statistical analysis paragraph in the past tense, in passive or active voice, with suggested software citations (R Core Team, IBM SPSS, StataCorp and more) and a checklist of what reviewers ask about statistics.
Which test fits which design?
| Design | Example | Main test | If assumptions fail | Effect size | Follow-up |
|---|---|---|---|---|---|
| Two independent groups | Treated vs control plots | Welch’s t-test (or Student’s) | Mann–Whitney U | Cohen’s d, Hedges’ g | None needed |
| Two related measurements | Before vs after, same patients | Paired t-test | Wilcoxon signed-rank | Cohen’s dz | None needed |
| Three or more independent groups | Five zinc rates | One-way ANOVA (Welch’s ANOVA if variances differ) | Kruskal–Wallis | ω², η² | Tukey HSD, Games–Howell, Dunnett; Dunn after Kruskal–Wallis |
| Two or more factors, or blocks | Zinc × irrigation in an RCBD | Factorial ANOVA or RCBD ANOVA | Transformation, or a generalised linear model | Partial η² | Estimated marginal means with Tukey adjustment |
| Repeated measures (3+ times) | Weeks 0, 4, 8 and 12 | Repeated-measures ANOVA (Greenhouse–Geisser) or a linear mixed model | Friedman | Partial η² | Bonferroni-adjusted pairwise tests |
| Two continuous variables | Leaf zinc and grain yield | Pearson’s r | Spearman’s ρ | r with 95% CI | None needed |
| Predicting a continuous outcome | Yield from N and P rate | Linear regression | Transformation or robust standard errors | b with 95% CI, R² | None needed |
| Predicting a yes/no outcome | Disease present or absent | Binary logistic regression | Exact or penalised logistic regression for rare events | Odds ratio with 95% CI | None needed |
| Two categorical variables | Treatment group × survival | Chi-square test of independence | Fisher’s exact test (expected count < 5) | Cramér’s V, odds ratio | Standardised residuals |
| Paired categorical data | Same patients, yes/no before and after | McNemar’s test | Exact McNemar test | Odds ratio | None needed |
Rules of thumb, not laws: the unit of analysis (plot, animal, patient) and how the data were collected decide the test. Not sure? Use the free which statistical test should I use? tool.
- PriceFree, no login or sign-up needed
- Your dataProcessed in your browser, never uploaded
- Works onPhone, tablet or computer, any modern browser
- Made byDr. Muhammad Aammar Tufail, Codanics
- Published22 November 2026
- Last updated22 November 2026
📚 How to cite this tool
This tool is free. If it helped your thesis, paper, report or assignment, please cite it. It keeps the tools free and helps other students find them.
Tufail, M. A. (2026). Statistical analysis paragraph generator [Online tool]. Codanics. Retrieved 22 November 2026, from https://codanics.com/tools/statistical-analysis-paragraph-generator/
In-text citation: (Tufail, 2026) · Narrative: Tufail (2026)
The access date is today’s date, filled in automatically. Author: Muhammad Aammar Tufail. Published 22 November 2026.
How to use the statistical analysis paragraph generator
- Pick your study design. The usual tests, assumption checks, post-hoc tests and effect sizes are ticked for you; change any of them to match what you actually did.
- Type your outcome(s) and your groups, factors or variables, so the sentences name them.
- Choose how you checked assumptions, what you did when they failed, your post-hoc tests and any correction for multiple comparisons.
- Choose the software, version and packages, the significance level, how you report the data, your effect sizes and how you handled missing data. Add a power analysis if you did one.
- Copy the paragraph (passive or active voice), add the suggested software citations to your reference list and work through the reviewer checklist.
What goes in a statistical analysis paragraph?
The statistical analysis paragraph is the last part of the methods section. It tells readers which test you used for which question, how you checked that the data met the test’s assumptions, how you handled multiple comparisons and missing data, what significance level you used and which software and version did the work. The test of a good paragraph is simple: another researcher with your data could repeat every analysis and get the same numbers.
| Element | Example sentence (past tense) | Why reviewers want it |
|---|---|---|
| Descriptive statistics | Data were summarised as mean ± SD. | Readers must know whether ± means SD or SE. |
| Assumption checks | Normality of the residuals was checked with the Shapiro–Wilk test and Q-Q plots. | Parametric tests are only valid when assumptions hold. |
| Tests | Differences among the five zinc rates were tested with one-way ANOVA. | Each result must be traceable to a test. |
| Fallback | When the assumptions were not met, the Kruskal–Wallis test was used instead. | Shows the choice was planned, not made after seeing p-values. |
| Post-hoc and corrections | Pairs of means were compared with Tukey’s HSD test. | Many comparisons inflate false positives. |
| Effect sizes | Effect sizes were reported as ω² with 95% CIs. | APA 7 and most journals ask for effect sizes and intervals. |
| Missing data | Two plots were lost to flooding and were excluded. | Missing data can bias results. |
| Significance level | All tests were two-tailed, with α = .05. | Defines what “significant” means in your paper. |
| Software and version | All analyses were performed in R version 4.4.1 (R Core Team, 2024). | Results can differ between versions and packages. |
Example statistical analysis paragraph
This is the paragraph the tool writes for the loaded example, a field trial with five zinc rates:
Passive or active voice?
Both are correct. The passive voice (“Data were analysed …”) is traditional in methods because the action matters more than who did it. Many journals, including Nature and the BMJ, now prefer the active voice (“We analysed …”), which is shorter and clearer. Use the voice your target journal uses, and keep it the same through the whole methods section. Methods are always in the past tense, because they describe what you did.
How to cite R, SPSS and other software
| Software | In the text | Reference |
|---|---|---|
| R | R version 4.4.1 (R Core Team, 2024) | R Core Team (2024). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/ |
| IBM SPSS Statistics | SPSS version 29.0 (IBM Corp., Armonk, NY, USA) | IBM Corp. Released 2022. IBM SPSS Statistics for Windows, Version 29.0. Armonk, NY: IBM Corp. |
| Stata | Stata 18 (StataCorp, College Station, TX, USA) | StataCorp. 2023. Stata Statistical Software: Release 18. College Station, TX: StataCorp LLC. |
| Python | Python 3.12 with SciPy (Virtanen et al., 2020) | Cite Python and the main libraries, e.g. SciPy: Virtanen et al. (2020), Nature Methods, 17, 261–272. |
In R, type citation() for R itself and citation("lme4") for a package; R prints the reference for the version you have installed. Cite the packages that did the main analysis, not every package you loaded.
What reviewers ask about statistics
- Is the test right for the design? Repeated measurements on the same plants or patients are not independent groups; several t-tests are not a substitute for ANOVA.
- Were the assumptions checked, and what happened when they failed?
- Which post-hoc test, and how were multiple comparisons controlled?
- How big is the effect? Report effect sizes and 95% confidence intervals, not only p-values.
- How was the sample size chosen, and how were missing data handled?
- Is ± SD or ± SE? SD describes the spread of the data; SE describes the precision of a mean.
The tool follows the APA Publication Manual (7th ed.) for statistics, the statistical items of the EQUATOR reporting guidelines (CONSORT, STROBE, ARRIVE) and the SAMPL guidelines for basic statistical reporting. Always check your journal’s guide for authors as well.
Frequently asked questions
How do I write the statistical analysis section of a research paper?
Say how data are summarised (for example mean ± SD), how assumptions were checked, which test was used for each comparison, what you did when assumptions failed, which post-hoc tests and corrections you used, which effect sizes you report, how missing data were handled, the significance level and the software with its version. This generator writes that paragraph from your choices.
Should the statistical analysis paragraph be in past or present tense?
Past tense, because it describes what you did: “Data were analysed with one-way ANOVA.” Use the present tense only for general facts, such as what a test assumes.
How do I cite R in my paper?
Write the version in the text, for example “R version 4.4.1 (R Core Team, 2024)”, and add the reference: R Core Team (2024). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/. Type citation() in R to get the reference for your version.
Do I need to report the Shapiro–Wilk test?
You should say how you checked normality. Many statisticians prefer Q-Q plots of the residuals, because with large samples the Shapiro–Wilk test flags trivial departures and with small samples it misses real ones. Stating both is common and safe.
What is the difference between mean ± SD and mean ± SE?
SD (standard deviation) shows how spread out the observations are. SE (standard error) shows how precisely the mean is estimated and gets smaller as the sample grows. Most journals ask for SD to describe data, and for 95% confidence intervals to show precision.
Which post-hoc test should I use after ANOVA?
Tukey’s HSD for all pairwise comparisons with similar variances, Games–Howell when variances differ, Dunnett’s test when every treatment is compared only with a control, and Dunn’s test after a significant Kruskal–Wallis test. Fisher’s LSD and Duncan’s test are still common in agriculture but control false positives poorly when there are many groups.
Does this tool run the statistics for me?
No. It writes the methods paragraph that describes the analysis you ran. To run tests, use the free Codanics calculators for t-tests, ANOVA, non-parametric tests and regression, or R, SPSS or Python.
Do I need to log in or create an account to use the statistical analysis paragraph generator?
No. There is no login, no sign-up and no email required. Open the page and start using it straight away, for free, on any device.
Free guides and courses from Codanics
More free tools
See all free Codanics tools → · Codanics on YouTube · WhatsApp channel