Describe your study instead
If your design doesn't fit the questions above (covariates, repeated measures, clustering), describe it and get a suggested analysis.
Uses an AI model via Vercel AI Gateway. It only sees what you send here; check anything you publish.
Quick reference
| Outcome | Two independent groups | Paired / repeated | 3+ groups | Relationship |
|---|---|---|---|---|
| Continuous, ~normal | Welch t-test | Paired t-test | One-way ANOVA | Pearson r, regression |
| Ordinal or skewed | Mann–Whitney U | Wilcoxon signed-rank | Kruskal–Wallis | Spearman ρ |
| Categorical | Two-proportion z, Fisher | McNemar | Chi-square | Chi-square |
Before collecting data, size the study with the power analysis calculator for the test you chose.
Frequently asked questions
How do I know which statistical test to use?
Three questions settle most cases: what you want to know (difference, relationship or estimate), what type of outcome you measured (continuous, categorical, ordinal), and how the groups relate (independent or paired, two or more).
When should I use a nonparametric test?
When the outcome is ordinal, or samples are small and clearly non-normal or contain outliers. With moderate to large samples, t-tests and ANOVA are robust to non-normality, so the parametric version is usually fine.
Which test compares three or more groups?
One-way ANOVA for independent groups with a continuous outcome, followed by Tukey HSD. The Kruskal–Wallis test (calculator on this site) is the rank-based alternative, and repeated-measures ANOVA is used when the same participants appear in every condition.
Do I need a test at all for a descriptive study?
Often not. If you only want to estimate a percentage or mean, report it with a confidence interval and plan the sample with a margin of error calculation.