Computer Generation of Poisson Deviates from Modified Normal Distributions

Samples from Poisson distributions of mean # _> 10 are generated by truncating suitable normal deviates and applying a correction with low probabdity. For p < 10, inversion is substituted. The method is accurate and it can cope with changing parameters p. Using efficient subprograms for generating uniform, exponential, alid normal deviates, the new algorithm is much faster than all competing methods.