Preliz Library

0.24.0 · active · verified Fri Apr 17

Preliz is a Python library for exploring and eliciting probability distributions. It provides a flexible and object-oriented framework for defining, manipulating, and visualizing various distributions, commonly used for setting priors in Bayesian inference workflows. The current version is 0.24.0, and it maintains an active development and release cadence.

Common errors

```
AttributeError: module 'preliz' has no attribute 'distributions'
```
cause Attempting to access a distribution via `preliz.distributions.DistName` after version `0.20.0`, when the `distributions` submodule was removed from direct access.

fix Update your import statements to directly import distributions from the top-level `preliz` module, e.g., `from preliz import DistName`.
```
TypeError: 'Normal' object cannot be interpreted as an integer
```
cause This or similar `TypeError` can occur when a `preliz` distribution object is passed to a function expecting a numerical value or a `scipy.stats` distribution, without proper conversion.

fix Ensure you are using the correct methods (`.pdf()`, `.cdf()`, `.rvs()`) or converting the `preliz` object to a `scipy.stats` object using `.to_scipy()` before passing it to external functions.
```
ImportError: cannot import name 'Normal' from 'preliz.distributions' (...path/to/preliz/distributions.py)
```
cause This error specifically indicates that an older import statement `from preliz.distributions import Normal` is being used with `preliz` versions `0.20.0` or later, where this import path is no longer valid.

fix Change the import statement to `from preliz import Normal` to import the distribution directly from the top-level `preliz` package.

Warnings

breaking Direct import paths for individual distributions changed in `0.20.0`. Previously, distributions like `Normal` were imported from `preliz.distributions`. Now, they are directly available from the top-level `preliz` module.
Fix: Update your import statements. For example, change `from preliz.distributions import Normal` to `from preliz import Normal`. A backward-compatible `from preliz import distributions` was added in `0.22.0`, but direct import is preferred.
gotcha Preliz distribution objects are not direct subclasses of `scipy.stats` distributions and cannot be used interchangeably with `scipy.stats` functions without explicit conversion. They offer their own `pdf()`, `cdf()`, `rvs()` methods.
Fix: Use the `to_scipy()` method available on `preliz` distribution objects to obtain a `scipy.stats` compatible distribution, or directly use the `preliz` object's native methods for calculations and sampling.
deprecated The plotting methods `plot_pdf()` and `plot_cdf()` on distribution objects have been deprecated in favor of a more unified `plot()` method.
Fix: Replace calls to `dist.plot_pdf()` or `dist.plot_cdf()` with `dist.plot()`. You can specify the desired plot type using the `plot_type` argument, e.g., `dist.plot(plot_type='pdf')`.

Install

pip install preliz Install stable version

Imports

Normal
wrong:
```
from preliz.distributions import Normal
```
correct:
```
from preliz import Normal
```
As of v0.20.0, distributions are directly available from the top-level 'preliz' module. The old path was deprecated.
preliz
```
import preliz
```
General import for accessing the library's features.

Quickstart

Demonstrates how to define a distribution, calculate its PDF, sample from it, and convert it to a scipy.stats compatible object.

import preliz
import numpy as np

# Define a Normal distribution
norm_dist = preliz.Normal(mu=0, sigma=1)

# Get PDF at a specific point
print(f"PDF at x=0: {norm_dist.pdf(0):.3f}")

# Sample from the distribution
samples = norm_dist.rvs(size=100)
print(f"Mean of 100 samples: {np.mean(samples):.2f}")
print(f"Std dev of 100 samples: {np.std(samples):.2f}")

# Access a scipy.stats compatible object (if needed)
scipy_norm = norm_dist.to_scipy()
print(f"Scipy PDF at x=0: {scipy_norm.pdf(0):.3f}")

view raw JSON →