STATISTICAL DECISION GUIDE
Start with the question, then choose the method. This guide narrows the choices from study design and outcome type. It recommends methods; it does not replace subject-matter judgment, protocol review, or assumption checks.
STEP 1
What is the main goal of your analysis? Choose the question that best matches the primary outcome and study objective. You can restart at any time.
Compare a numeric outcome between groups or conditions Examples: blood pressure, biomarker concentration, weight change. → Analyze counts, proportions, or categorical outcomes Examples: response yes/no, disease status, treatment category. → Measure association between variables Examples: age and blood pressure; symptom score and quality-of-life score. → Predict an outcome or adjust for covariates Examples: adjust a treatment comparison for age and baseline disease severity. → Analyze repeated or longitudinal measurements Examples: the same participant measured at baseline, week 4, and week 12. → Analyze time until an event Examples: time to death, relapse, hospital discharge, or device failure. → Assess agreement, reliability, or measurement consistency Examples: two radiologists reading the same scan; questionnaire items in a scale. → Explore many variables or identify groups/patterns Examples: reduce 30 biomarkers; identify patient subgroups. → Assess non-inferiority, equivalence, or superiority Examples: whether a new treatment is not worse than standard by a prespecified margin. → Evaluate diagnostic accuracy Examples: biomarker or imaging test against a reference disease diagnosis. → Check a numeric-data assumption before choosing a method Examples: whether paired differences or model residuals are reasonably normal; whether group variances look similar. →
Start over
Before choosing: confirm the outcome, independent versus paired/repeated observations, unit of analysis, missing-data plan, and whether the goal is estimation, hypothesis testing, prediction, or description.