¶ Description
Compare if two means are equal.
¶ Parameters
¶ Parameters tab
Parameters:
- How many rows (in percent) go to pin 0
- (optional) If a modality is unknown:
- Column name that contains the sample id
¶ Description tab
Parameters:
- Script name
- Short description
- Revision
- Decription
¶ Configuration tab
See dedicated page for more information.
¶ About
The splitStratified action button performs a stratified split of a dataset based on the distribution of a categorical (modal) variable. It routes a user-defined percentage of rows to the first output pin while ensuring the stratification is preserved across categories.
This action is useful in machine learning pipelines or experiments that require representative sampling across groups (e.g., class labels, education levels, etc.).
¶ Input
The action requires two input pins:
- Pin 0: Main data table to be split.
- Pin 1: A table with at least 2 columns including the stratification column (usually the modal variable).
The stratification column must be categorical and contain the same number of rows as Pin 0.
¶ Example
¶ Input Data (Pin 0)
| id | income | education |
|---|---|---|
| 1 | 50000 | Bachelor |
| 2 | 52000 | Master |
| 3 | 48000 | Bachelor |
| 4 | 51000 | PhD |
| 5 | 53000 | Master |
| 6 | 55000 | Bachelor |
¶ Stratification Column (Pin 1)
| _sample |
|---|
| 1 |
| 1 |
| 1 |
| 1 |
| 1 |
| 1 |
¶ Parameters
samplingParameter = 0.5idName = _sampleidUnknown = discard the row
¶ Output
Depending on the stratification and input data, 50% of the rows (approximately 3 rows) will be routed to the first output pin and the rest to the second — while preserving the proportion of each modal category (education in this case).
¶ 🛑 Errors & Troubleshooting
| Error Message | Cause | Solution |
|---|---|---|
No column on which to stratify |
Second input pin is empty or missing stratification column | Ensure Pin 1 contains the modal variable |
The table on the second pin must have 2 or more columns |
Pin 1 has only one column | Add the stratification column as well as ID |
Authorization required, but no protocol specified |
Internal runner issue | Re-run or ensure environment is authorized |
¶ Notes
- The stratified column must not be numeric with continuous values (e.g., floats); use categorical variables instead.
- Always ensure row counts match between both input pins.
¶ Screenshots
¶ Use Cases
- Train/test split for classification tasks
- Controlled sampling in survey or polling datasets
- Ensuring balanced distribution of classes in machine learning models