Quick answer: R Commander’s Statistics menu maps every classic test to its design: Means → t-tests and one-way ANOVA; Contingency tables → chi-square; Fit models → regression and ANOVA-family models; Non-Parametric tests → their rank-based twins. The skill isn’t clicking — it’s matching design to test first. This guide is that map, with a worked test per family. Setup and window tour: what is R Commander.
Design → test → menu path: the map
| Design | Test | Menu path |
|---|---|---|
| Two independent groups | Independent t-test (Welch if variances differ) | Statistics → Means → Independent-samples t-test |
| Same subjects twice | Paired t-test | Statistics → Means → Paired t-test |
| Three+ groups | One-way ANOVA + Tukey | Statistics → Means → One-way ANOVA |
| Two categorical variables | Chi-square / Fisher’s exact | Statistics → Contingency tables |
| Two continuous variables | Correlation / regression | Statistics → Fit models → Linear regression |
| Non-normal or ordinal data | Mann-Whitney, Wilcoxon, Kruskal-Wallis | Statistics → Non-Parametric tests |
One worked test: independent t-test
Two teaching methods, exam scores. Statistics → Means → Independent-samples t-test → outcome variable, group factor. If Levene’s test (output includes it) shows unequal variances, choose the Welch variant in the same dialog. Output gives t, df, p and CI — interpretation and effect sizes follow exactly the pattern in our jamovi t-test walkthrough; the underlying inference logic is in the hypothesis testing guide. Note what R Commander adds over other GUIs: the script window shows the exact call — t.test(score ~ method, data=mydata) — which you can edit and re-submit for variants the dialog doesn’t expose.
The three pre-test checks that save revisions
- Descriptives first (n, mean, SD per group — route: the descriptive statistics guide): impossible values and missingness show up here, not in the test.
- Shape check (histogram/QQ): badly skewed small samples route to Non-Parametric tests instead of forcing a t-test.
- Sample size per cell: chi-square needs expected counts ≥ 5 (Fisher’s exact for 2×2 shortfalls); t-tests with n ≥ 30 per group tolerate moderate skew.
Why all of this works — nulls, p-values, and what “significant” honestly means — is the short course in our p-value explainer; the biology-specific applications are in hypothesis testing in biology. For one-to-one coaching that turns menu knowledge into defensible analysis chapters, Ampersand Academy teaches R programming and biostatistics one-to-one.
Frequently asked questions
How do I run a t-test in R Commander?
Statistics menu, Means, then Independent-samples or Paired t-test. Choose the outcome and group variables, and the results appear with t, degrees of freedom, confidence interval and p-value – Levene’s test included for the independent case.
Where is chi-square in R Commander?
Statistics, Contingency tables, Enter and analyze two-way table – or statistics on a dataset when your data are one observation per row. Tick chi-square and expected counts in the dialog.
How do I do ANOVA in R Commander?
For a simple one-way design use Statistics, Means, One-way ANOVA. For factorial or more complex models use Statistics, Fit models, Linear model, then Models, Hypothesis tests for ANOVA tables and post hoc comparisons.
What if my data are not normally distributed?
Use the rank-based equivalents under Statistics, Non-Parametric tests: Mann-Whitney for two independent groups, Wilcoxon for paired, Kruskal-Wallis for three or more groups.
Can R Commander run regression?
Yes – Statistics, Fit models covers linear regression, logistic regression via the generalized linear model dialog, and more. The Models menu then provides diagnostics, predictions and coefficient tests.