Ziggurat normal random number generator ported to Python

NumPy installation is required. Python 3 compatible.

To install, run

python setup.py install

or you can install from PyPI:

pip install zignor

If you are running Windows and don't have MSVC compiler installed, a pre-built distribution on Python 3.5 is given in dist folder.

It will automatically import NumPy, so you can just use

import zignor
zignor.randn(10)

or

zignor.randn(3, 5)

to generate a matrix of 3 by 5.

This module includes two implementations. One is an slightly modified implementation of the original Ziggurat method (http://www.jstatsoft.org/article/downloadSuppFile/v005i08/rnorrexp.c). It uses SHR3 as the RNG for uniform distribution. This is the code behind zignor.randn() and zignor.rnor() method, and it is the faster of the two.

We also included a zignor.rexp() for generation exponential distribution.

The other implementation is the VIZIGNOR implementation by Doornik, J.A. It improves the original method and uses MWC8222 RNG to enable the method to pass some more tests. This method can be called by zignor.zignor().

1. G. Marsaglia and W.W. Tsang: The Ziggurat Method for Generating Random Variables. Journal of Statistical Software, vol. 5, no. 8, pp. 1–7, 2000.
2. Doornik, J.A. (2005), "An Improved Ziggurat Method to Generate Normal Random Samples", mimeo, Nuffield College, University of Oxford.