liminfoG*Power Reference
Free reference guide: G*Power Reference
About G*Power Reference
This G*Power Reference is a searchable cheat sheet for statistical power analysis and sample size calculation using G*Power software. It covers the five analysis types (a priori, post hoc, compromise, criterion, sensitivity), Cohen's effect size conventions (d, f, f-squared, w, r), and step-by-step input parameters for each test family.
The reference includes detailed guides for t-tests (independent, paired, one-sample, equivalence, Wilcoxon-Mann-Whitney), F-tests (one-way ANOVA, repeated measures, two-way ANOVA, ANCOVA, multiple regression R-squared, MANOVA), chi-square tests (goodness-of-fit, contingency table), z-tests (proportions, logistic regression, survival log-rank), and correlation analysis.
Designed for researchers, graduate students, biostatisticians, and clinical trial planners who need quick access to G*Power formulas, required sample sizes for common effect sizes, effect size conversion methods, multiple comparison corrections, and visualization features.
Key Features
- All five G*Power analysis types explained: a priori, post hoc, compromise, criterion, and sensitivity
- Cohen effect size conventions with small, medium, and large benchmarks for d, f, f-squared, w, and r
- Independent, paired, and one-sample t-test power analysis with sample size calculations
- ANOVA reference covering one-way, repeated measures, two-way, ANCOVA, and MANOVA
- Multiple regression power analysis for R-squared increase and R-squared deviation from zero
- Chi-square goodness-of-fit and contingency table power analysis with degrees of freedom
- Logistic regression, survival log-rank, and correlation bivariate analysis guides
- Effect size calculator conversions, Bonferroni/Sidak corrections, and X-Y plot visualization features
Frequently Asked Questions
What analysis types does G*Power support?
G*Power supports five analysis types: A priori (calculate required sample size from desired power), Post hoc (calculate achieved power from sample size), Compromise (analyze alpha/beta ratio), Criterion (derive alpha from power and sample size), and Sensitivity (determine detectable effect size from sample and power).
What are Cohen's effect size conventions?
Cohen defined small, medium, and large effect sizes for common measures: d = 0.2/0.5/0.8, f = 0.10/0.25/0.40, f-squared = 0.02/0.15/0.35, w = 0.10/0.30/0.50, and r = 0.10/0.30/0.50. These benchmarks help when no prior estimates of effect size are available for power calculations.
How do I calculate sample size for an independent samples t-test?
Select Test family t-test, then Means Two independent groups. Enter effect size d (e.g., 0.5 for medium), alpha level (typically 0.05), and desired power (typically 0.80). For d=0.5, alpha=0.05, and power=0.80, you need approximately n=64 per group, totaling N=128.
How do I perform power analysis for ANOVA in G*Power?
Select F test family and the specific ANOVA type (one-way, repeated measures, two-way). Enter effect size f (0.25 for medium), number of groups, alpha, and power. For a one-way ANOVA with f=0.25, 3 groups, alpha=0.05, and power=0.80, you need approximately n=53 per group (N=159 total).
How does the effect size calculator work?
G*Power includes a built-in Effect Size Calculator accessible via Determine then Effect size. It converts between measures: d to f (f = d/2 for 2 groups), eta-squared to f (f = sqrt(eta-squared/(1-eta-squared))), R-squared to f-squared (f-squared = R-squared/(1-R-squared)), and odds ratio to d.
What is the Bonferroni correction and how do I apply it?
The Bonferroni correction adjusts the significance level for multiple comparisons: alpha_adjusted = alpha / k (number of comparisons). For example, with 3 comparisons, use alpha = 0.05/3 = 0.017 in G*Power. The Sidak correction (1-(1-alpha)^(1/k)) is a less conservative alternative. Both increase the required sample size.
Can G*Power handle non-inferiority and equivalence testing?
Yes, select t-test family then Means Equivalence, which uses the TOST (Two One-Sided Tests) approach. Enter the equivalence margin (delta) and the actual expected difference as effect size. For non-inferiority, use a one-sided TOST with alpha=0.05 and desired power.
Is this G*Power reference free?
Yes, this G*Power cheat sheet is completely free with no registration required. It provides instant browser-based search across all test families and analysis types. It is part of liminfo.com's collection of free research and statistics reference tools.