A web-based sample size calculator for statistical power analysis, supporting T-Test, ANOVA, Paired T-Test, and Two Proportions calculations with interactive sensitivity analysis charts.
Simply visit the GPower Web Tool in a modern web browser. The tool will:
- Load the Python statistics engine (statsmodels and matplotlib)
- Present an interactive interface for entering parameters
- Calculate required sample sizes in real-time
- Display a sensitivity analysis chart showing N vs. Effect Size
-
Groups (T-Test / ANOVA): Calculate required sample sizes for comparing group means
- T-Test: For comparing 2 independent groups using Cohen's d effect size
- ANOVA: For comparing 3+ groups using Cohen's f effect size
-
Paired T-Test (Before/After): Calculate required sample sizes for dependent samples
- For comparing paired measurements (e.g., before vs. after intervention)
- Uses Cohen's d effect size with correlation adjustment
- Adjusts for reduced variance in paired designs: d_adj = d / √(1 - ρ)
-
Two Proportions: Calculate required sample sizes for comparing two independent proportions (dichotomous outcomes)
- Uses Cohen's h effect size (displayed in real-time)
- Ideal for clinical trials comparing treatment vs. control success rates
A key feature that transforms GPower from a single-number calculator into a visual decision-making tool with dual visualization:
- Dual Chart Layout: Side-by-side charts showing N vs Effect Size and N vs Power
- N vs Effect Size Chart: Shows how sample size requirements change across different effect sizes with your fixed Power and Alpha
- N vs Power Chart: Shows how sample size requirements change as you vary statistical power with your fixed Effect Size and Alpha
- Cohen's Conventions: Vertical lines mark Small, Medium, and Large effect size thresholds
- Standard Power Line: Marks the conventional Power = 0.80 threshold
- Current Position Indicator: Red dots show your current parameters and required N
- Real-time Updates: Charts regenerate instantly as you adjust parameters
- Professional Output: Clean, publication-ready dual charts using Matplotlib
-
Dropout Rate Adjustment: Account for expected participant dropouts using the formula:
$$\text{Required Sample} = \frac{\text{Theoretical Sample}}{1 - \text{Dropout Rate}}$$ - Interactive Interface: Real-time calculations with intuitive sliders and input controls
- Effect Size Presets: Quick selection of small, medium, and large effect sizes based on Cohen's conventions
- Mobile Responsive: Works seamlessly on both desktop and mobile devices
- No Installation Required: Runs entirely in your browser using Pyodide (Python in WebAssembly)
- Number of Groups: 2 for T-test, 3+ for ANOVA
- Effect Size: Expected magnitude of the effect (Cohen's d for T-test, Cohen's f for ANOVA)
- Statistical Power: Probability of detecting an effect if it exists (typically 0.80)
- Significance Level (alpha): Risk of Type I error (typically 0.05)
- Expected Dropout Rate: Percentage of participants expected to drop out (0-50%)
- Effect Size (Cohen's d): Standardized difference between paired measurements
- Expected Correlation (ρ): Correlation between paired observations (0.01-0.99, default 0.50)
- Statistical Power: Probability of detecting an effect if it exists (typically 0.80)
- Significance Level (alpha): Risk of Type I error (typically 0.05)
- Expected Dropout Rate: Percentage of participants expected to drop out (0-50%)
- Baseline Rate / Control Group: Expected proportion in the control group (e.g., 0.50)
- Expected Treatment Rate: Expected proportion in the treatment group (e.g., 0.70)
- Cohen's h: Effect size calculated automatically from the two proportions
- Statistical Power: Probability of detecting an effect if it exists (typically 0.80)
- Significance Level (alpha): Risk of Type I error (typically 0.05)
- Expected Dropout Rate: Percentage of participants expected to drop out (0-50%)
The GPower tool's calculations are based on established statistical methods and implemented using a widely-cited, open-source Python library. We recommend citing the following works in your research to validate the methodological approach:
The principles of power analysis and the conventions for effect sizes (Cohen's d, f, and h) are based on the seminal work in this field:
Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Lawrence Erlbaum Associates.
The real-time calculations rely on the statistical models implemented in the Python statsmodels library, which is run in the browser via Pyodide. This library provides the power functions for Means (T-Test/ANOVA) and Proportions (Normal Approximation).
Seabold, S., & Perktold, J. (2010). Statsmodels: Econometric and statistical modeling with python. Proceedings of the 9th Python in Science Conference.
To ensure you use the most current and accurate citation for this software, please use the citation feature provided by GitHub.
Click the "Cite this repository" button (usually visible on the right sidebar of the GitHub repository page) to find various formats (e.g., APA, BibTeX) generated automatically from the CITATION.cff file.
This project is open source and available under the MIT License.