People think choosing a statistical test is an exact science. It isn’t. Often, there’s no single definitive right answer. The tests are like screwdrivers of slightly different sizes… one might fit a little better than another, but several will get the job done in many non-academic environments. The danger isn’t picking the second-best option. It’s getting so stuck on finding the perfect test that you never analyse anything at all.
This selector walks you through a few plain language questions about your data – what type it is, how many groups you’re comparing, what you’re trying to find out – and points you to a sensible place to start. No textbook, no statistics degree needed.
The selector covers the tests most teams meet in real improvement work:
- Comparing groups: t-tests, ANOVA, Mann-Whitney, Kruskal-Wallis, chi-square and more
- Relationships between variables: correlation and regression
- Data type: continuous, count, or categorical
- Paired vs independent samples: before/after data, matched groups, and unpaired comparisons
- Parametric vs non-parametric: when your data doesn’t meet normality assumptions
How to use your result: treat the selector’s answer as a sensible first option, not a verdict. If two tests are close, check the assumptions and pick the one you can defend. A good enough test, applied and interpreted properly, beats the mathematically perfect test you never run – and if your data is unstable or your measure is weak, no test will save it. Go look at the process first.
Statistical test selector
Answer the questions below to find the right statistical test for your data. Click any step in the trail to go back.
