01 One variable 02 Two variables / groups 03 Multiple groups / variables
Compare one sample with a reference value or distribution
Data and research question Design and conditions Methods to consider
01 Numerical outcome versus a specified mean Population SD known, or a justified normal approximation
Population SD unknown and estimated from the sample
Ordinal data, or a location analysis based on signs or ranks
02 Categorical outcome versus specified probabilities Binary outcome; test whether the event probability equals a specified value
Three or more categories; observed counts versus specified proportions
First distinguish independent, paired, and association designs
Data and research question Design and conditions Methods to consider
01 Compare two numerical outcomes Repeated or matched observations; analyze within-pair differences
Independent participants; compare means or distributional locations
02 Two categorical variables or a 2 × 2 table Independent observations; test whether the categorical variables are independent
Paired binary outcomes or before-and-after measurements
Estimate an odds ratio in case-control data or relative risk in cohort data
03 Association between numerical or ordinal variables Linear relationship; describe correlation or explain Y with X
Monotonic but not necessarily linear, or rank-based analysis
Test a correlation or compare correlations from independent samples
04 Two time-to-event groups Censored observations are present; compare the full survival curves
Distinguish one factor, multiple factors, multiple outcomes, and multiple predictors
Data and research question Design and conditions Methods to consider
01 Numerical or ordinal outcome across three or more independent groups Compare group means with a common within-group variance
Compare group means when variances differ
Rank-based comparison when ordinal data or a parametric model is unsuitable
02 Multiple factors or multiple outcome variables One numerical outcome with two categorical factors and their interaction
Analyze several correlated numerical outcomes jointly
03 One outcome with multiple predictors Numerical outcome; estimate each coefficient while adjusting for other predictors
This navigator locates candidate methods; it does not replace study-design judgment. Check the null hypothesis, independence, distributional conditions, and small-sample calculation in the linked note.
Biostatistics Study NotesFrom concepts and derivations to interpretation