https://github.com/thequackdaddy/tweedie

Tweedie family density estimation in python
https://github.com/thequackdaddy/tweedie

distribution glm insurance statistics tweedie

Last synced: about 2 months ago
JSON representation

Tweedie family density estimation in python

• Host: GitHub
• URL: https://github.com/thequackdaddy/tweedie
• Owner: thequackdaddy
• License: bsd-2-clause
• Created: 2017-04-25T01:44:52.000Z (over 7 years ago)
• Default Branch: master
• Last Pushed: 2023-11-17T17:51:07.000Z (about 1 year ago)
• Last Synced: 2024-10-31T08:42:08.026Z (about 2 months ago)
• Topics: distribution, glm, insurance, statistics, tweedie
• Language: Python
• Homepage:
• Size: 111 KB
• Stars: 27
• Watchers: 3
• Forks: 15
• Open Issues: 2
• Metadata Files:
  • Readme: README.rst
  • License: LICENSE.txt

Awesome Lists containing this project

README

        
tweedie is a Python library implementing scipy's ``rv_continuous`` class

for the Tweedie family. The Tweedie family is a member of the exponential

dispersion model family and is commonly used in the insurance indsutry

to model claim amounts for insurance policies (exposure).

The main focus of this package is the compound-Poisson behavior,

specifically where ``1 < p < 2``. However, it should be possible to

calculate the distribution for all the possible values of p.

.. image:: https://app.travis-ci.com/thequackdaddy/tweedie.svg?branch=master

   :target: https://app.travis-ci.com/thequackdaddy/tweedie

.. image:: http://codecov.io/github/thequackdaddy/tweedie/coverage.svg?branch=master

   :target: http://codecov.io/github/thequackdaddy/tweedie?branch=master

Documentation:

  Some of the functions are well documented. Others, not so well. In future

  versions, I might try to publish.

Downloads:

  https://github.com/thequackdaddy/tweedie

Dependencies:

  * Python (2.7, or 3.4+)

  * numpy

  * scipy

Optional dependencies:

  * pytest: needed to run tests

  * tox: needed to run tests against multiple python/scipy versions

Install:

  To run tests against multiple python/scipy versions, install the package using ``pip install -e .[dev]``.

Running the tests:

To test against the current python version and scipy ``pytest ./tweedie``

To test against multiple python versions and scipy versions ``tox -p all``.

If you run in to issues with performance, just use ``tox``.

Code and bug tracker:

  https://github.com/thequackdaddy/tweedie

License:

  2-clause BSD, see LICENSE.txt for details.