Conseiller de méthodes statistiques
"Which statistical method should I use for my data?" — answer a few questions and the advisor recommends a statistical method and takes you straight to the matching online tool. Covers common scenarios: group comparison, correlation/regression, diagnostic tests, survival analysis, and agreement/reliability. For method selection only; make the final call in light of your study design.
What is the goal of your study?
Mode d'emploi et méthodologie
Can this advisor make the decision for me?
It gives a 'recommended direction' for common situations to help you quickly settle on a suitable method and go straight to the tool. The actual choice also depends on the study design, data distribution, sample size, and meaning of the variables; for complex or atypical designs, consult a statistician.
How do I judge whether data are normal?
Look at a histogram/Q-Q plot or run a normality test; with larger samples, t-tests/ANOVA are fairly robust to departures from normality, while for small samples or clear skew, prefer non-parametric methods. When unsure, take the 'skewed/unsure' branch — non-parametric is the safer choice.
What if the method I need isn't among the recommended tools?
This site keeps adding tools. If a method you need (e.g. Poisson regression, mixed-effects models) isn't live yet, use the closest method for now, or tell us via the feedback link.
Why are there both parametric and non-parametric recommendations for the same kind of question?
Parametric methods (t-test, ANOVA, Pearson) have higher power but require assumptions like normality/equal variance; non-parametric methods (Mann-Whitney, Kruskal-Wallis, Spearman) have looser assumptions and are more robust to skew and outliers. Choose based on whether your data meet the assumptions.
Collez le lien dans Slack, Teams, X ou LinkedIn — l’aperçu utilise la carte Open Graph de cette page.