When to use this
Use a procedure that matches the outcome scale, group structure and independence or pairing in the study design.
Why it matters
Different tests estimate different standard errors and reference distributions; a convenient test is not automatically an appropriate one.
Step-by-step instructions
1. Confirm independent groups or paired observations.
2. Select the outcome and grouping variables.
3. Choose Welch when equal variances are not justified.
4. Review diagnostics and effect sizes.
5. Interpret intervals and design limitations, not only p-values.
Expected outputs
Group summaries, test statistic, degrees of freedom, p-value, confidence interval, effect sizes and relevant diagnostics.
How to interpret the results
The p-value is conditional on the null model and assumptions. Report uncertainty and effect size alongside the test.
Common mistakes
Do not use independent tests for paired data, infer pairwise differences from an omnibus ANOVA alone, or use chi-square with severely sparse expected counts without review.
Was this article helpful?
Your response helps the methodology team improve this documentation.