In recent years, simulation has become an important methodology for applied decision analysis under uncertainty. A typical simulation effort requires generating a set of possibly correlated and non-normal random variables, using information regarding their underlying joint probability density function contained in a presumably random sample. A few techniques have been suggested to accomplish this task, but most have not met ·the requirement of being correct and efficient from the statistical point of view. This study proposes a multivariate hyperbolic sine probability density function as a basis to develop an efficient and theoretically consistent approach for generating correlated, non-normal random variables.