A Relation Between Decision Making Penalty and Simulation Sample Size for Inventory Systems

The paper contains results from experiments in which two inventory models, with known analytical solutions have been simulated to learn about simulation. The author suggests a relation between the simulation sample size and the economic penalty for an incorrect decision from inaccuracy in the simulated information. The relation is W̄ = A/n^k, where W̄ is the average penalty in dollars, n is the sample size, and A and k are parameters obtained from a regression analysis. A series of results are reported with values given for A and K. The values of A vary monotonically with some of the inventory parameters. The values of k have moderate variability. It is hoped that the ideas suggested and the result presented may be helpful to others conducting experiments to learn more about simulation.