Techniques for Multivariate Simulation from Mixed Marginal Distributions with Application to Whole-Farm Revenue Simulation

Alternative techniques for representing dependencies among variables in multivariate simulation are discussed and compared in the context of rating a whole-farm insurance product. A procedure by lman and Conover (IC) that is common in actuarial applications is compared to a new technique detailed by Phoon, Quek, and Huang (PQHl. Results suggest that rates derived from the IC procedure may be inaccurate because the procedure produces biased estimates of correlation between simulated variables. This situation is improved with the PQH procedure.