AMDG On 2/10/2011 9:25 AM, Stephan Kitchovitch wrote:

I was looking through the Math/Statistical Distributions and Random boost libraries to see if they provide an efficient way to generate random numbers from the multinomial distribution. I have read that one way of achieving this is to generate from binomial distributions sequentially, although I am uncertain as to how efficient this approach would be.

The binomial_distribution implementation in the current release is very slow for large n. I've written a replacement that should be reasonably fast. It will be released in 1.47. While it's probably possible to generate multinomial variates faster using a direct algorithm, you probably can't get more than a factor of 2, possibly less.

Are there any alternative efficient approaches I could use? Does boost provide any other libraries that may help with this?

If anyone could recommend another reliable random number generator that does generate from a multinomial dist it would be great!

In Christ, Steven Watanabe