Results
Interpretation
Descriptive statistics
Group comparisons
| Group A | Group B | Mean difference | Median difference | Variance ratio | Cliff's delta |
|---|
Statistical tests
Detected outliers
| Row | Group | Subgroup | Original value | Transformed value | Method |
|---|
Configuration summary
Formula used
A violin plot mirrors a kernel density estimate around each category. Wider sections indicate more observations near that value. The inner box summarizes quartiles and the median.
Automatic bandwidth balances detail and smoothness using the available sample. Manual bandwidth provides direct control over density smoothing. Smaller values reveal detail but may amplify noise.
How to use
- Paste values or upload a CSV file.
- Select value, group, and optional subgroup columns.
- Choose density, inner-box, point, and transformation settings.
- Configure outlier rules and statistical comparisons.
- Generate the plot and review statistics.
- Export the chart, tables, cleaned data, or PDF.
Example data
| value | group | subgroup |
|---|---|---|
| 0.81 | Model A | Validation |
| 0.84 | Model A | Validation |
| 0.76 | Model B | Validation |
| 0.79 | Model B | Validation |
Machine learning uses
Compare feature distributions across target classes. Inspect residuals, confidence scores, losses, runtimes, anomaly scores, and cross-validation results. Evaluate shifts between training, validation, and test datasets.
Violin plots reveal spread, skewness, multiple peaks, and unequal sample sizes. They retain more distribution detail than standard box plots. Statistical tables support careful model comparison.
Frequently asked questions
What does violin width represent?
Width represents estimated data density. Wider regions contain more nearby observations.
How is a violin plot different from a box plot?
A box plot summarizes quartiles. A violin plot also shows the distribution shape.
What does bandwidth control?
Bandwidth controls density smoothing. Small bandwidths show detail, while large bandwidths produce smoother shapes.
Can I compare several machine learning models?
Yes. Assign each model name to the group column and provide repeated scores.
What is a split violin plot?
A split violin places two subgroup distributions on opposite sides of the same category.
How are outliers detected?
Use the IQR rule or z-score threshold. You may include, exclude, or simply list them.
Should I remove outliers?
Not automatically. Investigate their cause and compare results before excluding them.
Which transformation should I use?
Use transformations only when they match your analytical goal. Log transforms require positive values.
What do multiple peaks mean?
Multiple peaks may indicate subpopulations, mixed classes, or distinct generating processes.