https://github.com/jabulente/kruskall-wallis-test
This repository contain project that provides a reusable Python function to perform the Kruskal-Wallis H-test across multiple continuous variables, grouped by a categorical feature
https://github.com/jabulente/kruskall-wallis-test
data-analysis data-science eda hypothesis-tests kruskal-wallis kruskals-algorithm scipy-stats statistics
Last synced: 3 months ago
JSON representation
This repository contain project that provides a reusable Python function to perform the Kruskal-Wallis H-test across multiple continuous variables, grouped by a categorical feature
- Host: GitHub
- URL: https://github.com/jabulente/kruskall-wallis-test
- Owner: Jabulente
- License: mit
- Created: 2025-07-16T09:42:44.000Z (3 months ago)
- Default Branch: main
- Last Pushed: 2025-07-16T13:08:14.000Z (3 months ago)
- Last Synced: 2025-07-17T13:32:08.757Z (3 months ago)
- Topics: data-analysis, data-science, eda, hypothesis-tests, kruskal-wallis, kruskals-algorithm, scipy-stats, statistics
- Language: Jupyter Notebook
- Homepage:
- Size: 36.1 KB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
Kruskal-Wallis Test for Multiple Variables and Group Comparisons
This project provides a reusable Python function to perform the **Kruskal-Wallis H-test** across **multiple continuous variables**, grouped by a categorical feature. It returns a **clean summary DataFrame** with test statistics, p-values, and significance indicators, making it easy to evaluate whether medians differ significantly between groups for each variable.
## 1. Purpose
The Kruskal-Wallis test is a **non-parametric alternative to one-way ANOVA**. It is used when:
- You have **3 or more independent groups**
- Your data **violates ANOVA assumptions** (e.g., normality, homogeneity of variance)
- Your data is **ordinal or continuous but not normally distributed**
This script simplifies applying this test across **many variables at once**, saving time and boosting productivity in exploratory and inferential analysis.
## 2. Features
- Accepts a pandas DataFrame and a grouping column
- Automatically applies the Kruskal-Wallis test to all other numeric variables
- Outputs a summary table including:
- Variable name
- Kruskal-Wallis test statistic
- p-value
- Significance status (`p < 0.05`)
- Ready for integration in statistical reports or dashboards
## 3. How to Use
```python
from scipy.stats import kruskal
import pandas as pd
import numpy as np
def kruskall_wallis(df, group_columns: str, numerical_columns: list = None):
if numerical_columns is None:
numerical_columns = df.select_dtypes(include=[np.number]).columns.tolist()
for g in group_columns:
if g in numerical_columns:
numerical_columns.remove(g)
results = []
for group_column in group_columns:
for column in numerical_columns:
# Create a list of samples grouped by group_column
groups = [group[column].dropna().values for name, group in df.groupby(group_column)]
stats, p_value = kruskal(*groups)
interpretation = '✔' if p_value < 0.05 else '✖'
results.append({
'Group': group_column,
'Variables': column,
'Kruskal-Wallis Statistic': stats,
'P-value': p_value,
'Significant (α<0.05)': interpretation
})
return pd.DataFrame(results)
```
## 4. 📂 Example Dataset
```
df = pd.DataFrame({
'Group 1': ['Ashura', 'Ashura', 'Ashura', 'Barack', 'Barack', 'Barack', 'Colins', 'Colins', 'Colins'],
'Group 2': ['Orenge', 'Orenge', 'Orenge', 'Banana', 'Banana', 'Banana', 'Carott', 'Carott', 'Carott'],
'Group 3': ['Alpha', 'Alpha', 'Alpha', 'Bravo', 'Bravo', 'Bravo', 'Eagle', 'Eagle', 'Eagle'],
'Variable 1': [12, 14, 13, 15, 16, 14, 10, 9, 11],
'Variable 2': [7, 6, 7, 8, 9, 10, 5, 6, 5],
'Variable 3': [20, 21, 19, 23, 22, 21, 18, 17, 19],
'Variable 4': [124, 145, 137, 150, 163, 148, 180, 90, 111],
'Variable 5': [70, 66, 75, 80, 92, 100, 56, 64, 56],
'Variable 6': [2, 2, 1, 2, 2, 2, 1, 1, 1]
})
groups_column = ['Group 1', 'Group 2', 'Group 3']
results = kruskal_test_all_variables(df, 'Group')
print(results)
```
## 5. Sample Output
Variable Kruskal-Wallis Statistic p-value Significant (p<0.05)
| | Group | Variables | Kruskal-Wallis Statistic | P-value | Significant (α<0.05) |
|---:|:--------|:------------|---------------------------:|----------:|:-----------------------|
| 0 | Group 1 | Variable 1 | 6.88 | 0.032 | ✔ |
| 1 | Group 1 | Variable 2 | 6.997 | 0.03 | ✔ |
| 2 | Group 1 | Variable 3 | 6.531 | 0.038 | ✔ |
| 3 | Group 1 | Variable 4 | 2.4 | 0.301 | ✖ |
| 4 | Group 1 | Variable 5 | 7.261 | 0.027 | ✔ |
| 5 | Group 1 | Variable 6 | 5.6 | 0.061 | ✖ |
| 6 | Group 2 | Variable 1 | 6.88 | 0.032 | ✔ |
| 7 | Group 2 | Variable 2 | 6.997 | 0.03 | ✔ |
| 8 | Group 2 | Variable 3 | 6.531 | 0.038 | ✔ |
| 9 | Group 2 | Variable 4 | 2.4 | 0.301 | ✖ |
| 10 | Group 2 | Variable 5 | 7.261 | 0.027 | ✔ |
| 11 | Group 2 | Variable 6 | 5.6 | 0.061 | ✖ |
| 12 | Group 3 | Variable 1 | 6.88 | 0.032 | ✔ |
| 13 | Group 3 | Variable 2 | 6.997 | 0.03 | ✔ |
| 14 | Group 3 | Variable 3 | 6.531 | 0.038 | ✔ |
| 15 | Group 3 | Variable 4 | 2.4 | 0.301 | ✖ |
| 16 | Group 3 | Variable 5 | 7.261 | 0.027 | ✔ |
| 17 | Group 3 | Variable 6 | 5.6 | 0.061 | ✖ |
# 6. 📈 Applications
- Agricultural and biological experiments
- Social science and behavioral research
- Market and product group analysis
- Education and clinical trial comparisons
## 7. 📥 Requirements
- Python 3.x
- pandas
- scipy
```
pip install pandas scipy
```
## 8. 🤝 Contributing
Feel free to fork this repo, contribute improvements, or suggest additional features such as post-hoc Dunn tests, visualization tools, or effect size calculation.
⭐️ If You Found This Useful
- Leave a star 🌟
- Share feedback
- Fork and adapt for your project
- Mention the project in your work!
## 9. License
MIT License