GSD

Reference implementation of generalised score distribution in python

This library provides a reference implementation of gsd probabilities for correctness and efficient implementation of samples and log_probabilities in jax.

Citations

Theoretical derivation of GSD is described in the following papers.

@Article{Cmiel2023,
author={{\'{C}}miel, Bogdan
and Nawa{\l}a, Jakub
and Janowski, Lucjan
and Rusek, Krzysztof},
title={Generalised score distribution: underdispersed continuation of the beta-binomial distribution},
journal={Statistical Papers},
year={2023},
month={Feb},
day={09},
issn={1613-9798},
doi={10.1007/s00362-023-01398-0},
url={https://doi.org/10.1007/s00362-023-01398-0}
}

@ARTICLE{gsdnawala,
  author={Nawała, Jakub and Janowski, Lucjan and Ćmiel, Bogdan and Rusek, Krzysztof and Pérez, Pablo},
  journal={IEEE Transactions on Multimedia}, 
  title={Generalized Score Distribution: A Two-Parameter Discrete Distribution Accurately Describing Responses From Quality of Experience Subjective Experiments}, 
  year={2022},
  volume={},
  number={},
  pages={1-15},
  doi={10.1109/TMM.2022.3205444}
  }

If you decide to apply the concepts presented or base on the provided code, please do refer our related paper.

Installation

You can install gsd via pip:

pip install ref_gsd

Note that you install ref_gsd but import gsd e.g.

import gsd

gsd.fit_moments([2, 8, 2, 0, 0.])

Development

To develop and modify gsd, you need to install hatch, a tool for Python packaging and dependency management.

To enter a virtual environment for testing or debugging, you can run:

hatch shell

Running tests

Gsd uses unitest for testing. To run the tests, use the following command:

hatch run test 

Standalone estimator

You can quickly estimate GSD parameters from a command line interface

python3 -m gsd -c 1 2 3 4 5

psi=3.6667 rho=0.6000

Acknowledgments

Development of this software has received funding from the Norwegian Financial Mechanism 2014-2021 under project- 2019/34/H/ST6/00599.