Group Comparison & ANOVA
Comparing two or more groups on a continuous outcome is the most common task in experimental research. Licklider walks you through the full path — from normality check and variance homogeneity to test selection, effect size, and journal-ready figure — while enforcing every mandatory disclosure.
Verified execution paths (parametric-first): Independent t-test · Welch's t-test ·
One-way ANOVA · Two-way ANOVA · Post-hoc (Tukey, Holm, Bonferroni, etc.) · Effect size (Cohen's d, η², ω²).
Guidance-only in the current product: Mann–Whitney / Kruskal–Wallis / Wilcoxon-style rank tests — see
Non-Parametric Alternatives.
Repeated-measures ANOVA / LMM: coverage varies; check method pages and Known Limitations.
When to Use This Section
Use Group Comparison & ANOVA workflows when:
- You have a continuous outcome variable and one or more categorical grouping factors
- You want to test whether group means (or distributions) differ significantly
- You need to report effect size, CI, and post-hoc results alongside significance annotations
- Your design includes repeated measurements, paired observations, or factorial combinations
Available Scenarios
Two-Group Comparison
Two-group Welch/t-test path (verified). Normality check → test selection → box plot + strip overlay. Rank-based follow-up: guidance page only.
One-Way ANOVA
Four dose groups with Holm-corrected post-hoc — pharmacologist with planned contrasts and CLD annotation.
Two-Way ANOVA
Treatment × sex interaction — neuroscientist with factorial design and interaction plot disclosure.
Repeated Measures
Repeated-measures ANOVA and LMM — longitudinal behavioral experiment with 5 time points and sphericity correction.
Test Selection Guide
Licklider's automatic selection and engine execution prioritize parametric routes (Welch, ANOVA, post-hoc). The UI may surface rank-based method labels when assumptions suggest them, but Mann–Whitney, Wilcoxon, Kruskal–Wallis, and Friedman are not verified end-to-end through the public stats engine — see Non-Parametric Alternatives.
- 2 groups, independent, normal: Welch's t-test (default over Student's t)
- 2 groups, independent, non-normal: Selection may recommend a rank-based label; execution remains parametric-first — use the guidance page to plan outside-Licklider analysis if needed
- 2 groups, paired: Paired t-test when a subject/block ID column is specified (verified path)
- 3+ groups, independent: One-way ANOVA → Holm or Tukey post-hoc (verified)
- 3+ groups, non-normal / unequal n: Review Non-Parametric Alternatives; no dedicated rank-based omnibus path is verified in-product
- 2 factors: Two-way ANOVA with interaction term (where supported)
- Repeated measures: See method pages; not all RM/LMM paths are fully verified — Known Limitations