Asymptotically Optimal Allocation of Simulation Experiments in Discrete Stochastic Optimization

Approximate solutions for discrete stochastic optimization problems are often obtained via simulation. It is reasonable to complement these solutions by confidence regions for the argmin-set. We address the question, how a certain total number of random draws should be distributed among the set of alternatives. We propose a one-step allocation rule which turns out to be asymptotically optimal in the case of normal errors for two goals: To minimize the costs caused by using only an approximate solution and to minimize the expected size of the confidence sets