Which statistical test should I use?

This interactive decision tree helps you choose an appropriate basic parametric test. Assume:

Continuous outcome variable
Approximately normally distributed residuals
Independent observations where required

How to use this tool

Answer each question based on your study design, not the result you hope to find. The tool guides you to a suitable statistical test and highlights key assumptions.

Parametric vs non-parametric

Parametric tests

These tests (e.g., t-tests, ANOVA, Pearson correlation) assume the data follow a specific distribution, usually normality of residuals. They are generally more powerful when assumptions are met.

Non-parametric tests

These tests (e.g., Mann–Whitney U, Wilcoxon, Friedman, Spearman’s rho) do not rely on normality assumptions and often use ranks. They are useful for ordinal data, small samples, or clear assumption violations.

Key terms

Between-subjects: Different participants in each group.

Within-subjects: The same participants measured multiple times.

Sphericity: Equality of variances of differences between repeated measures conditions.

Levene’s test: Checks equality of variances between groups.

Effect size: Quantifies the magnitude of an effect, beyond statistical significance.