PoissonBinomial

PoissonBinomial - An R package for efficient computation of generalized binomial distributions

Brief description

During the development of the R package FDX, it was decided to move the Poisson binomial distribution functions to a separate package in order to make the more efficient algorithms available to other R users. These are significantly faster than implementations in existing packages. Our package was designed to include the usual functions for point probabilities, cumulative probabilities, quantiles, and random numbers. A total of ten algorithms are available for each of these functions, including four exact procedures and six approximations. Later, methods for the generalized Poisson binomial distribution were added. The completed package was published on the Comprehensive R Archive Network (CRAN) for distribution.

Project data

Project manager

Florian Junge

Contact persons at DISO

Time period

October 2019 to July 2021

Project history

  • October 2019: Implementation of the first algorithms within the FDX package
  • November 2019: Decision for further development in a separate package
  • December 2019: Preliminary completion
  • January 2020: Release of the first version 1.0 on CRAN
  • February 2020: Version 1.0.1; bug fixes
  • February 2020: Version 1.0.2; bug fixes, performance improvements
  • April 2020: Version 1.1.0; addition of the generalized Poisson binomial distribution, further bug fixes
  • May 2020: Version 1.1.1; bug fixes
  • August 2020: Version 1.1.2; numerical improvements, further bug fixes
  • September 2020: Version 1.1.3; further numerical improvements
  • November 2020: Version 1.2.0; performance improvements, more random number generators
  • December 2020: Version 1.2.1; minor bug fixes for Solaris systems
  • March 2021: Version 1.2.2/1.2.3; major performance improvements for exact distribution functions
  • July 2021: Version 1.2.4; performance improvements and bug fixes for quantile functions

Back

 

Menu