Abstract
An automatic method is developed for the computer generation of random variables with a characteristic function satisfying certain regularity conditions. The method is based upon a generalization of the rejection method and exploits the duality between densities and their Fourier transforms. It takes finite time almost surely, does not use approximations or inversions, and does not require explicit knowledge of the characteristic function (only its computability is assumed—hence the adjective “automatic”). As a by-product, we show how the sum of n independent random variables with common density f can be generated in time essentially independent of n, at least when its characteristic function satisfies the above mentioned regularity conditions.

This publication has 10 references indexed in Scilit: