Multivariate input modeling with Johnson distributions

This paper introduces a new method for multivariate simulation input modeling based on the Johnson translation system of probability distributions. This technique matches the first four marginal moments and the correlation structure of a given set of sample data, allowing computationally efficient parameter estimation and random-vector generation. Applications of the technique in ergonomics and production scheduling are discussed. The proposed method is compared to traditional multivariate input-modeling techniques based on the Johnson translation system.