Pykelihood is a Python package for statistical analysis designed to give more flexibility to likelihood-based inference than is possible with scipy.stats. Distributions are designed from an Object Oriented Programming (OOP) point of view. In particular, this package allows the fitting of complex distributions to a dataset, add trends of different forms in the parameters of the target distribution, condition the log-likelihood with any form of penalty, and profile parameters of the model based on the chosen likelihood's sensitivity.
Main features include:
- use any scipy.stats distribution, or make your own,
- fit distributions of arbitrary complexity to your data,
- add trends of different forms in the parameters of any distribution,
- condition the log-likelihood with any form of penalty,
- profile parameters with a penalised log-likelihood,
- more to come...
The toy examples presented in the documentation may seem simple, but the package is highly practical in many cases. It allows for fitting trends, conditioning log-likelihoods with custom penalty functions, and reparameterizing distributions like the GEV in terms of return levels (which is essential in the field of extreme event attribution, for instance). It also simplifies tasks such as building profile likelihood-based confidence intervals, making it a flexible and efficient tool for statistical modeling.
For detailed documentation, please visit the official documentation.
pip install pykelihoodgit clone https://www.github.com/OpheliaMiralles/pykelihoodor
gh repo clone OpheliaMiralles/pykelihoodPoetry is used to manage pykelihood's dependencies and build system. To install
Poetry, you can refer to the installation instructions, but it boils
down to running:
curl -sSL https://raw.githubusercontent.com/python-poetry/poetry/master/get-poetry.py | python
To configure your environment to work on pykelihood, run:
git clone https://www.github.com/OpheliaMiralles/pykelihood # or any other clone method cd pykelihood poetry install
This will create a virtual environment for the project and install the required dependencies. To activate the virtual
environment, be sure to run poetry shell prior to executing any code.
We also use the pre-commit library which adds git hooks to the repository. These must be installed with:
pre-commit install
Some parts of the code base use the matplotlib and hawkeslib package, but are for now not required to run most of the code, including the tests.
Tests are run using pytest. To run all tests, navigate to the root folder or the
tests folder and type pytest.