Methods
Use this overview to decide whether your question belongs in Methods, understand what Methods does not cover, and jump to the leaf page that matches your analysis question.
Use this section when your main question is which statistical method fits your study design, outcome type, and inferential goal.
This section is for choosing an analysis method family, not for free-form figure selection, general quality auditing, or data setup. If you primarily need a chart, assumption checks, or study-setup guidance, start with Figures and Visualization, Quality Checks, or Study Setup instead.
Methods groups Licklider's analysis guidance by research question: group differences, outcome-predictor modeling, categorical association, time-to-event outcomes, power and sample size planning, robust or resampling-based alternatives, and specialized outcome types such as counts, proportions, ordinal outcomes, and compositional data.
Start here
Use the table below to jump to the leaf page that most closely matches your question. If more than one row seems plausible, choose the row defined by your outcome type first, then refine by design.
| If your main question is... | Start here |
|---|---|
| Do 2 groups differ on a continuous outcome? | t-Test |
| Do 3 or more groups differ on one factor? | One-Way ANOVA and Post Hoc |
| Do groups differ across 2 factors or an interaction? | Two-Way ANOVA and Post Hoc |
| Do the same subjects contribute measurements under 3+ conditions? | Repeated Measures ANOVA |
| Do groups differ across one between-subjects factor and one within-subjects factor? | Mixed ANOVA |
| Do I need a rank-based or non-parametric group comparison? | Non-Parametric Alternatives |
| How does a continuous outcome change with predictors? | Linear Regression (OLS) |
| How are Pearson and Spearman correlations reported on scatter or regression figures? | Correlation Analysis |
| How do I model a binary outcome or evaluate classification with ROC/AUC? | Logistic Regression and AUC/ROC |
| Do I need a curved dose-response model such as IC50 or 4PL? | Non-linear Regression and IC50/4PL |
| Do I need random effects, clustering, or partially dependent observations? | GLMM: Gaussian and Binomial |
| Does the same subject contribute repeated measurements? | Repeated Measures and Mixed Models |
| Are 2 categorical variables associated in a contingency table? | Chi-Square Test |
| Is the sample small enough that I need an exact contingency-table test? | Fisher Exact Test |
| When does an event happen, and how does time-to-event differ between groups? | Kaplan-Meier Analysis |
| How do covariates relate to hazard over time? | Cox Proportional Hazards Regression |
| How large should my study be before I collect data? | Power Analysis and Sample Size Calculation |
| What effect size would count as practically meaningful before analysis? | Minimal Effect of Interest |
| Do I need bootstrap intervals, permutation tests, or a Bayes factor supplement? | Bootstrap Confidence Intervals, Permutation Tests, or Bayes Factor Supplement |
| Is my outcome a count, proportion, bounded response, ordinal score, or composition? | Count Data Models, Proportion and Bounded Response Data, Ordinal Outcome Analysis, or Compositional Data Analysis |
What belongs in Methods
Choose a Methods page when you need to answer a research question with a named statistical procedure or model. Common examples include Welch t-tests, one-way or two-way ANOVA, Mann-Whitney U, linear regression, logistic regression, ROC/AUC, Kaplan-Meier curves, Cox regression, permutation tests, bootstrap confidence intervals, and count or ordinal models.
What does not belong here
Methods helps you choose the analysis family; it is not the place to choose a figure style, audit dataset quality in the abstract, or define your data contract before analysis.
Category map
- Group Comparison: Use when the question is whether groups differ on an outcome.
- Regression and Modeling: Use when the question is how an outcome changes with one or more predictors.
- Categorical and Association: Use when the question is whether categorical variables are associated.
- Survival Analysis: Use when the outcome is time to event, censoring, or hazard over follow-up.
- Power and Sample Size: Use before data collection to plan sample size or define a minimal effect of interest.
- Robust Resampling and Bayesian: Use when you need bootstrap, permutation, robust alternatives, or Bayes factor support around a core analysis.
- Specialized Outcome Types: Use when the outcome scale itself drives the method choice, such as counts, proportions, bounded outcomes, ordinal responses, or compositions.
Related
- Related links are registered in frontmatter and rendered below this page.