Choosing a statistical method
Start with your question and study design. Explore suitable methods, understand their assumptions, and open a calculator when the design fits.
Browse all methodsDecision matrix
Compare study scenarios. Select a suggestion to see its rationale and assumptions.
| Purpose | Study / question | Suggested method |
|---|---|---|
| Compare groups or a reference | Numeric measurements, such as blood pressure → One sample against a reference → Mean against a reference | |
| Compare groups or a reference | Numeric measurements, such as blood pressure → One sample against a reference → Symmetric location against a reference | |
| Compare groups or a reference | Numeric measurements, such as blood pressure → One sample against a reference → Median / direction against a reference | |
| Compare groups or a reference | Numeric measurements, such as blood pressure → Two groups or conditions → Different independent subjects → Difference in means (Welch) | |
| Compare groups or a reference | Numeric measurements, such as blood pressure → Two groups or conditions → Different independent subjects → Distribution / rank comparison | |
| Compare groups or a reference | Numeric measurements, such as blood pressure → Two groups or conditions → Different independent subjects → Difference in means with exchangeable labels | |
| Compare groups or a reference | Numeric measurements, such as blood pressure → Two groups or conditions → Matched pairs / the same subjects twice → Mean change | |
| Compare groups or a reference | Numeric measurements, such as blood pressure → Two groups or conditions → Matched pairs / the same subjects twice → Symmetric location using signed ranks | |
| Compare groups or a reference | Numeric measurements, such as blood pressure → Two groups or conditions → Matched pairs / the same subjects twice → Direction only / sign test | |
| Compare groups or a reference | Numeric measurements, such as blood pressure → Two groups or conditions → Additional repeated or clustered structure → Continuous / Gaussian, random intercept adequate | |
| Compare groups or a reference | Numeric measurements, such as blood pressure → Two groups or conditions → Additional repeated or clustered structure → Binary, count, or other non-Gaussian outcome | Guidance only |
| Compare groups or a reference | Numeric measurements, such as blood pressure → Two groups or conditions → Additional repeated or clustered structure → Gaussian hierarchical Bayesian analysis | |
| Compare groups or a reference | Numeric measurements, such as blood pressure → Two groups or conditions → Additional repeated or clustered structure → I’m unsure about the dependence structure | Guidance only |
| Compare groups or a reference | Numeric measurements, such as blood pressure → Two groups or conditions → I’m not sure | Guidance only |
| Compare groups or a reference | Numeric measurements, such as blood pressure → Three or more / factorial design → One factor, independent groups → Means, allowing unequal variances | |
| Compare groups or a reference | Numeric measurements, such as blood pressure → Three or more / factorial design → One factor, independent groups → Means with a justified common variance | |
| Compare groups or a reference | Numeric measurements, such as blood pressure → Three or more / factorial design → One factor, independent groups → Rank distributions | |
| Compare groups or a reference | Numeric measurements, such as blood pressure → Three or more / factorial design → One factor, independent groups → Planned pairwise / control comparisons → All pairs, common-variance means | |
| Compare groups or a reference | Numeric measurements, such as blood pressure → Three or more / factorial design → One factor, independent groups → Planned pairwise / control comparisons → All pairs, unequal-variance means | |
| Compare groups or a reference | Numeric measurements, such as blood pressure → Three or more / factorial design → One factor, independent groups → Planned pairwise / control comparisons → Independent-group rank comparisons (Holm) | |
| Compare groups or a reference | Numeric measurements, such as blood pressure → Three or more / factorial design → One factor, independent groups → Planned pairwise / control comparisons → Several means against one control | |
| Compare groups or a reference | Numeric measurements, such as blood pressure → Three or more / factorial design → Same subjects in several conditions → One within-subject factor, mean comparison | |
| Compare groups or a reference | Numeric measurements, such as blood pressure → Three or more / factorial design → Same subjects in several conditions → Rank comparison across conditions | |
| Compare groups or a reference | Numeric measurements, such as blood pressure → Three or more / factorial design → Same subjects in several conditions → Repeated outcomes plus between-subject predictors | |
| Compare groups or a reference | Numeric measurements, such as blood pressure → Three or more / factorial design → Multiple factors / adjusted comparison → Two between-subjects factors | |
| Compare groups or a reference | Numeric measurements, such as blood pressure → Three or more / factorial design → Multiple factors / adjusted comparison → Three between-subjects factors | |
| Compare groups or a reference | Numeric measurements, such as blood pressure → Three or more / factorial design → Multiple factors / adjusted comparison → One group factor plus one covariate | |
| Compare groups or a reference | Numeric measurements, such as blood pressure → Three or more / factorial design → Multiple factors / adjusted comparison → Repeated outcomes plus predictors | |
| Compare groups or a reference | Numeric measurements, such as blood pressure → Three or more / factorial design → Other grouped continuous observations | |
| Compare groups or a reference | Numeric measurements, such as blood pressure → Three or more / factorial design → Several continuous outcomes jointly | |
| Compare groups or a reference | Numeric measurements, such as blood pressure → Three or more / factorial design → I’m not sure | Guidance only |
| Compare groups or a reference | Ordered ratings, such as mild / moderate / severe → Two independent groups | |
| Compare groups or a reference | Ordered ratings, such as mild / moderate / severe → Two paired conditions, meaningful symmetric differences | |
| Compare groups or a reference | Ordered ratings, such as mild / moderate / severe → Three or more independent groups | |
| Compare groups or a reference | Ordered ratings, such as mild / moderate / severe → Three or more repeated conditions | |
| Compare groups or a reference | Ordered ratings, such as mild / moderate / severe → Adjust for predictors | |
| Compare groups or a reference | Ordered ratings, such as mild / moderate / severe → Unsure about pairing or rating differences | Guidance only |
| Compare groups or a reference | Two categories, such as event / no event → One independent sample vs expected proportion → Binary proportion, exact test | |
| Compare groups or a reference | Two categories, such as event / no event → One independent sample vs expected proportion → Binary proportion, large-sample approximation | |
| Compare groups or a reference | Two categories, such as event / no event → One independent sample vs expected proportion → Several category counts vs expected counts | |
| Compare groups or a reference | Two categories, such as event / no event → Independent groups → Contingency table with adequate expected counts | |
| Compare groups or a reference | Two categories, such as event / no event → Independent groups → Sparse 2×2 count table | |
| Compare groups or a reference | Two categories, such as event / no event → Independent groups → Sparse table larger than 2×2 | Guidance only |
| Compare groups or a reference | Two categories, such as event / no event → Independent groups → Two proportions with adequate expected counts | |
| Compare groups or a reference | Two categories, such as event / no event → Same subjects twice / matched pairs | |
| Compare groups or a reference | Two categories, such as event / no event → Same subjects in three or more conditions | |
| Compare groups or a reference | Two categories, such as event / no event → Adjust for predictors | |
| Compare groups or a reference | Two categories, such as event / no event → Clustered observations / repeated predictors | Guidance only |
| Compare groups or a reference | Several unordered categories → One variable vs an expected distribution → Binary proportion, exact test | |
| Compare groups or a reference | Several unordered categories → One variable vs an expected distribution → Binary proportion, large-sample approximation | |
| Compare groups or a reference | Several unordered categories → One variable vs an expected distribution → Several category counts vs expected counts | |
| Compare groups or a reference | Several unordered categories → Association in independent groups → Contingency table with adequate expected counts | |
| Compare groups or a reference | Several unordered categories → Association in independent groups → Sparse 2×2 count table | |
| Compare groups or a reference | Several unordered categories → Association in independent groups → Sparse table larger than 2×2 | Guidance only |
| Compare groups or a reference | Several unordered categories → Association in independent groups → Two proportions with adequate expected counts | |
| Compare groups or a reference | Several unordered categories → Paired categories with more than two levels | Guidance only |
| Compare groups or a reference | Several unordered categories → Adjust for predictors | |
| Compare groups or a reference | Time until an event, with possible censoring → Estimate survival curves | |
| Compare groups or a reference | Time until an event, with possible censoring → Unadjusted comparison of survival groups | |
| Compare groups or a reference | Time until an event, with possible censoring → Cox regression with covariates | |
| Compare groups or a reference | Time until an event, with possible censoring → Parametric survival / extrapolation | |
| Compare groups or a reference | Time until an event, with possible censoring → Competing risks | |
| Compare groups or a reference | Variability rather than location → One normal population variance vs reference | |
| Compare groups or a reference | Variability rather than location → Two independent normal population variances | |
| Compare groups or a reference | Variability rather than location → Compare groups with median-centered Levene | |
| Compare groups or a reference | I’m not sure | Guidance only |
| Examine a relationship | Two numeric variables → Linear association | |
| Examine a relationship | Two numeric variables → Monotonic rank association | |
| Examine a relationship | Two numeric variables → Rank concordance with tied ranks | |
| Examine a relationship | Two numeric variables → One predictor and a numeric outcome | |
| Examine a relationship | Ordered variables / monotonic relationship | |
| Examine a relationship | Two categorical variables → Contingency table with adequate expected counts | |
| Examine a relationship | Two categorical variables → Sparse 2×2 count table | |
| Examine a relationship | Two categorical variables → Sparse table larger than 2×2 | Guidance only |
| Examine a relationship | Two categorical variables → Two proportions with adequate expected counts | |
| Examine a relationship | Predict an outcome or adjust for covariates → Continuous numeric outcome → One predictor, independent observations | |
| Examine a relationship | Predict an outcome or adjust for covariates → Continuous numeric outcome → Several predictors, independent observations | |
| Examine a relationship | Predict an outcome or adjust for covariates → Continuous numeric outcome → Repeated / clustered observations | |
| Examine a relationship | Predict an outcome or adjust for covariates → Binary outcome | |
| Examine a relationship | Predict an outcome or adjust for covariates → Ordered categories | |
| Examine a relationship | Predict an outcome or adjust for covariates → Unordered categories (3+) | |
| Examine a relationship | Predict an outcome or adjust for covariates → Event count or rate → Poisson mean–variance relationship | |
| Examine a relationship | Predict an outcome or adjust for covariates → Event count or rate → Overdispersion consistent with an NB2 model | |
| Examine a relationship | Predict an outcome or adjust for covariates → Event count or rate → Repeated / clustered counts | Guidance only |
| Examine a relationship | Predict an outcome or adjust for covariates → Time to event with censoring | |
| Examine a relationship | Predict an outcome or adjust for covariates → Repeated / clustered outcome → Continuous / Gaussian, random intercept adequate | |
| Examine a relationship | Predict an outcome or adjust for covariates → Repeated / clustered outcome → Binary, count, or other non-Gaussian outcome | Guidance only |
| Examine a relationship | Predict an outcome or adjust for covariates → Repeated / clustered outcome → Gaussian hierarchical Bayesian analysis | |
| Examine a relationship | Predict an outcome or adjust for covariates → Repeated / clustered outcome → I’m unsure about the dependence structure | Guidance only |
| Model or predict an outcome | Continuous numeric outcome → One predictor, independent observations | |
| Model or predict an outcome | Continuous numeric outcome → Several predictors, independent observations | |
| Model or predict an outcome | Continuous numeric outcome → Repeated / clustered observations | |
| Model or predict an outcome | Binary outcome | |
| Model or predict an outcome | Ordered categories | |
| Model or predict an outcome | Unordered categories (3+) | |
| Model or predict an outcome | Event count or rate → Poisson mean–variance relationship | |
| Model or predict an outcome | Event count or rate → Overdispersion consistent with an NB2 model | |
| Model or predict an outcome | Event count or rate → Repeated / clustered counts | Guidance only |
| Model or predict an outcome | Time to event with censoring | |
| Model or predict an outcome | Repeated / clustered outcome → Continuous / Gaussian, random intercept adequate | |
| Model or predict an outcome | Repeated / clustered outcome → Binary, count, or other non-Gaussian outcome | Guidance only |
| Model or predict an outcome | Repeated / clustered outcome → Gaussian hierarchical Bayesian analysis | |
| Model or predict an outcome | Repeated / clustered outcome → I’m unsure about the dependence structure | Guidance only |
| Describe data or estimate a quantity | Numeric summaries | |
| Describe data or estimate a quantity | One mean and confidence interval | |
| Describe data or estimate a quantity | Difference between independent means | |
| Describe data or estimate a quantity | One proportion and confidence interval | |
| Describe data or estimate a quantity | Difference between independent proportions | |
| Describe data or estimate a quantity | Categorical rater agreement | |
| Describe data or estimate a quantity | Questionnaire internal consistency | |
| Explore structure or classify observations | Reduce numeric variables → Components summarizing observed variation | |
| Explore structure or classify observations | Reduce numeric variables → Explore latent factors | |
| Explore structure or classify observations | Reduce numeric variables → Evaluate a prespecified factor model | |
| Explore structure or classify observations | Discover groups of observations → Numeric data, partition into k groups | |
| Explore structure or classify observations | Discover groups of observations → Numeric data, nested Ward hierarchy | |
| Explore structure or classify observations | Discover groups of observations → Mixed numeric / categorical data or two-step | Guidance only |
| Explore structure or classify observations | Classify known group labels | |
| Explore structure or classify observations | Relate two sets of numeric variables | |
| Plan a study or use a specialist method | Sample size and study design → Common mean, proportion, ANOVA or regression design | |
| Plan a study or use a specialist method | Sample size and study design → Noninferiority / equivalence | |
| Plan a study or use a specialist method | Sample size and study design → Cluster randomized trial | |
| Plan a study or use a specialist method | Sample size and study design → Survival trial | |
| Plan a study or use a specialist method | Sample size and study design → Group sequential / futility | |
| Plan a study or use a specialist method | Sample size and study design → Adaptive sample-size re-estimation | |
| Plan a study or use a specialist method | Sample size and study design → Bayesian trial simulation | |
| Plan a study or use a specialist method | Survival analysis → Estimate survival curves | |
| Plan a study or use a specialist method | Survival analysis → Unadjusted comparison of survival groups | |
| Plan a study or use a specialist method | Survival analysis → Cox regression with covariates | |
| Plan a study or use a specialist method | Survival analysis → Parametric survival / extrapolation | |
| Plan a study or use a specialist method | Survival analysis → Competing risks | |
| Plan a study or use a specialist method | Combine study evidence → Pool study effects | |
| Plan a study or use a specialist method | Combine study evidence → Explore study-level moderators | |
| Plan a study or use a specialist method | Observational causal analysis → Propensity matching | |
| Plan a study or use a specialist method | Observational causal analysis → Propensity weighting | |
| Plan a study or use a specialist method | Observational causal analysis → Augmented inverse probability weighting | |
| Plan a study or use a specialist method | Observational causal analysis → Difference-in-differences | |
| Plan a study or use a specialist method | Health economics / patient simulation → Decision tree rollback | |
| Plan a study or use a specialist method | Health economics / patient simulation → Cohort Markov / cost effectiveness | |
| Plan a study or use a specialist method | Health economics / patient simulation → Patient microsimulation | |
| Plan a study or use a specialist method | Health economics / patient simulation → Probabilistic sensitivity analysis | |
| Plan a study or use a specialist method | Health economics / patient simulation → One-way sensitivity analysis | |
| Plan a study or use a specialist method | Health economics / patient simulation → Budget impact | |
| Plan a study or use a specialist method | Health economics / patient simulation → Patient flow / resources | |
| Plan a study or use a specialist method | Experimental design / process statistics → Factorial experimental design | |
| Plan a study or use a specialist method | Experimental design / process statistics → Response surface | |
| Plan a study or use a specialist method | Experimental design / process statistics → Process capability | |
| Plan a study or use a specialist method | Experimental design / process statistics → Control charts | |
| Plan a study or use a specialist method | Pharmacology / bioequivalence → Noncompartmental analysis | |
| Plan a study or use a specialist method | Pharmacology / bioequivalence → Compartmental PK | |
| Plan a study or use a specialist method | Pharmacology / bioequivalence → Exposure–response | |
| Plan a study or use a specialist method | Pharmacology / bioequivalence → Crossover bioequivalence | |
| Plan a study or use a specialist method | Bayesian modeling → Bayesian regression | |
| Plan a study or use a specialist method | Bayesian modeling → Gaussian hierarchical model |