Generation of Inverse Gaussian Variates with Given Sample Mean and Dispersion

SUMMARY A method is described for the computer generation of a random sample of inverse Gaussian variates with given sample mean and measure of dispersion. The method can be extended to produce samples from other distributions. This way of generating random samples is particularly useful when applying variance reduction methods such as those of antithetic variates or stratified sampling in a Monte Carlo simulation. For illustration, an application is described to an estimation problem connected with the Weibull or extreme-value distribution.