Risk analysis : Frequentist and Bayesians unite!

Both deterministic and random simulation models may have unknown values for their parameters and input variables. Simulation analysts may therefore assume that these values have a statistical distribution (based on expert opinions or pastdata). Thisassumptionresultsinso-calledRiskAnalysis or Uncertainty Analysis. Frequentists use crude Monte Carlo, Latin Hypercube Sampling, or some other method to sample values from this assumed input distribution; next, the simulation model transforms these input values into output values; repeating this sampling many times gives an Estimated Distribution Function for the outputs; this function provides an estimate of the probability of a specific ‘disaster’. Software is abundant for this analysis. Bayesians use more specific input distributions; e.g., a conjugate prior distribution. They also obtain simulation data, but next they compute the posterior output distribution. Frequentists and Bayesians should unite to exchange experiences and theories.