Hyppo: Independence Testing

0.5.2 · active · verified Fri Apr 17

Hyppo is a comprehensive independence testing package in Python, currently at version 0.5.2. It provides a wide array of statistical tests for evaluating independence, k-sample, and goodness-of-fit problems. The library has an active development cycle, with major releases bringing significant new features and bug fixes, typically every 12-16 months, complemented by more frequent patch releases.

Common errors

```
ModuleNotFoundError: No module named 'autograd'
```
cause The `autograd` library, a mandatory dependency for Hyppo, is not installed or not accessible in your Python environment.

fix Install `autograd` explicitly: `pip install autograd`.
```
RuntimeError: Python version must be >= 3.8.0 to use hyppo>=0.5.0
```
cause You are attempting to use Hyppo version 0.5.0 or newer with a Python version older than 3.8.

fix Upgrade your Python interpreter to version 3.8 or newer, or downgrade Hyppo to version 0.4.x or older: `pip install 'hyppo<0.5.0'`.
```
AttributeError: 'MyCustomKernel' object has no attribute 'gradient'
```
cause If `MyCustomKernel` inherits from `DifferentiableKernel` and you are using `hyppo` v0.5.2 or newer, the `autograd` fallback for gradients was removed. You must now provide your own `gradient` method.

fix Implement the `gradient` method within your `MyCustomKernel` class to define how its gradients are computed.

Warnings

breaking Hyppo versions 0.5.0 and newer require Python 3.8 or higher. Users on older Python versions (3.6 or 3.7) will encounter errors if they upgrade past 0.4.x.
Fix: Upgrade your Python environment to 3.8+ or pin `hyppo` to a version less than 0.5.0 (e.g., `pip install 'hyppo<0.5.0'`).
breaking Starting from v0.5.2, the `DifferentiableKernel` class no longer provides an automatic `autograd` fallback for gradients. If you are implementing custom differentiable kernels, you must now manually define the `gradient` method.
Fix: For custom `DifferentiableKernel` subclasses, ensure you explicitly implement the `gradient` method with the desired differentiation logic. Do not rely on `autograd` to implicitly provide it.
gotcha The `autograd` library is a core dependency for `hyppo`. While `pip install hyppo` typically handles this, if you encounter `ModuleNotFoundError` related to `autograd`, it might be missing or conflict with your environment.
Fix: Manually install `autograd`: `pip install autograd`. Ensure your Python environment is clean or use a virtual environment.

Install

pip install hyppo Install latest version

Imports

Hsic
```
from hyppo.independence import Hsic
```
Dcorr
```
from hyppo.independence import Dcorr
```
Energy
```
from hyppo.ksample import Energy
```

Quickstart

This example demonstrates how to use the Hilbert-Schmidt Independence Criterion (HSIC) to test the independence between two sets of data, `x` and `y`.

import numpy as np
from hyppo.independence import Hsic

# Generate some data
x = np.random.normal(0, 1, size=(100, 1))
y = x + np.random.normal(0, 1, size=(100, 1))

# Initialize and run the Hsic test
test = Hsic()
stat, pvalue = test.test(x, y)

print(f"Test Statistic: {stat:.4f}, P-value: {pvalue:.4f}")

view raw JSON →