Prefer to click through? → Interactive chooser Fits one page — A4 or US Letter.
StatsCapybara · Cheat sheet

Which Test Should I Use?

Find your outcome and design in the left column; the test is on the right. Sections (like 2.1) point to the interactive lesson.

① Compare means · a numeric outcome

Your designTest§
One group vs. a known valueOne-sample t-test1.11
Same people, measured twicePaired t-test1.11
Two separate groupsIndependent t-test (Welch)1.12
3+ groups, one factorOne-way ANOVA2.1
3+ groups, two crossed factorsFactorial ANOVA2.3
Same people, 3+ conditionsRepeated-measures ANOVA2.4
Adjust for a baseline covariateANCOVA3.4
Several outcomes tested jointlyMANOVA3.11

② Assumptions broken? · swap to ranks

Parametric testRank-based swap§
Independent t-testMann–Whitney U2.6
Paired / one-sample tWilcoxon signed-rank2.6
One-way ANOVAKruskal–Wallis2.6
Repeated-measures ANOVAFriedman2.6

Reach for these when the outcome is ordinal, or badly skewed / outlier-ridden in a small sample. With a large n, the parametric tests are robust anyway.

③ Relationships & association

Your variablesTest§
Two numeric, roughly linearPearson r2.8
Two numeric, monotone / outliersSpearman ρ2.8
Two categoricalChi-square test2.7

④ Predict / model an outcome

What you're predictingModel§
A number, one predictorSimple linear regression2.9
A number, several predictorsMultiple regression3.1
Yes / no (binary)Logistic regression3.7
A count (0, 1, 2, …)Poisson / NB (GLM)4.4
Nested / clustered / repeatedMixed (multilevel) model4.5
Time until an event, with dropoutsSurvival (Kaplan–Meier / Cox)4.8
Combining many studiesMeta-analysis4.10
Max accuracy, not a hypothesis testMachine-learning modelsML 4.1

Before you trust the answer

1 · Check the assumptions. Every test above has them — see the Assumption Checks sheet, or §2.5.
2 · Pick one test, not many. Running several and keeping whatever hits p < .05 manufactures false positives — the garden of forking paths. Decide before you look.
3 · A p-value needs an effect size. Significant ≠ large — always report one beside it (§1.13).
4 · One numeric + one categorical variable is just a group comparison in disguise → box ①.