Discrete random variate generation using uniformization (Q1071468)

The paper describes a method (called uniformization) of generation of a discrete random variable with almost arbitrary distribution function F. The generation is based on the first passage time of an associated stochastic process. The paper contains proofs of the needed results, formulae for some special cases of distributions and results of simulation experiments. The method works for a wide class of distributions better than the straightforward method based on the inverse of distribution functions. It is slower for ''standard'' distributions (Poisson, binomial, etc.) than the best known methods. Some extensions of the method are investigated. The extensions seem to be very promising even for ''standard'' distributions.