Use Cases Overview Group Comparison & ANOVA ↳ Two-Group Comparison ↳ One-Way ANOVA ↳ Two-Way ANOVA ↳ Repeated Measures Regression & Curve Fitting ↳ Linear Regression ↳ Logistic Regression & AUC ↳ Dose-Response & IC50 ↳ Mixed Models (GLMM) Survival Analysis ↳ Kaplan-Meier Workflow ↳ Cox Regression Workflow Normality & Power ↳ Pre-Study Power Calculation ↳ Normality Testing Categorical & Association ↳ Chi-Square Test ↳ Fisher's Exact Test ↳ Odds Ratio & Risk Life Sciences Applications ↳ Flow Cytometry ↳ Gene Expression / Omics ↳ Animal Study Workflow ↳ Clinical Trial Endpoints

Normality & Power

Two of the most common analytical questions researchers ask: "How many samples do I need?" (before the experiment) and "Is my data normal enough to use parametric tests?" (after data collection). Licklider provides defensible, journal-ready answers to both.

Section covers: Shapiro-Wilk · Kolmogorov-Smirnov · D'Agostino-Pearson · Q-Q plots · Power analysis (G*Power-equivalent) · Effect size input (Cohen's d, η², f) · Alpha and beta specification · Sample size output with assumptions declared

When to Use This Section

Use Normality & Power workflows when:

  • You need to justify a sample size before starting an experiment (grant application, ethics committee)
  • You want to check whether your data distribution is consistent with parametric test assumptions
  • You need to report power and α level in your Methods section as required by journals
  • A reviewer or editor asks you to justify why you used (or did not use) a parametric test

Available Scenarios

Key Concepts

  • Power (1-β): Probability of detecting a true effect. Typically set to 0.80 (80%) or 0.90 (90%).
  • α (Type I error): Probability of a false positive. Typically set to 0.05.
  • Effect size: The magnitude of the difference you expect to detect. Must be estimated from pilot data, literature, or clinically meaningful differences — not set arbitrarily.
  • Normality testing: With small n (<10 per group), Shapiro-Wilk has low power — failing to reject normality does not prove normality. With large n (>100), trivial deviations from normality will be significant — clinical significance matters more than statistical significance.
  • Parametric vs. non-parametric: The decision is based on the distribution of residuals (in ANOVA) or data (in t-tests), not group means. Visual inspection (Q-Q plot, histogram) alongside a formal test provides the most defensible justification.