10. Conclusion

MCMC is a surprisingly difficult and bug-prone algorithm to implement by hand. We find PyMC makes it much easier and less stressful. PyMC also makes our work more dynamic; getting hand-coded MCMC’s working used to be so much work that we were reluctant to change anything, but with PyMC changing models is much easier.

We welcome new contributors at all levels. If you would like to contribute new code, improve documentation, share your results or provide ideas for new features, please introduce yourself on our mailing list. Our wiki page. also hosts a number of tutorials and examples from users that could give you some ideas. We have taken great care to make the code easy to extend, whether by adding new database backends, step methods or entirely new sampling algorithms.

11. Acknowledgements

The authors would like to thank several users of PyMC who have been particularly helpful during the development of the 2.0 release. In alphabetical order, these are Mike Conroy, Abraham Flaxman, J. Miguel Marin, Aaron MacNeil, Nick Matsakis, John Salvatier, Andrew Straw and Thomas Wiecki.

Anand Patil’s work on PyMC has been supported since 2008 by the Malaria Atlas Project, principally funded by the Wellcome Trust.

David Huard’s early work on PyMC was supported by a scholarship from the Natural Sciences and Engineering Research Council of Canada.

Project Versions

Table Of Contents

Previous topic

9. Probability distributions

Next topic

12. Appendix: Markov Chain Monte Carlo

This Page