> The best approach that I can recall to this problem is to use a modified DeMoivre distribution (see Kohler 2000 - [www.demographic-research.org/articles/volume/3/8](http://www.demographic-research.org/articles/volume/3/8))
