Linking statistical estimation and decision making through simulation

Models that are built to help make decisions usually involve input parameters, which need to be estimated statistically using data. However, submitting these estimated parameters directly to the model may result in biased decisions because the estimated parameters are biased or the model is nonlinear. We propose a new parameter estimator called Simulation-Based Inverse Estimator (SBIE) to link the statistical estimation and decision making together. The linkage is achieved by simulating the model and adjusting the estimated parameters such that the adjusted parameters can adapt to the specific model. We prove that SBIE can provide us with consistent and unbiased decisions under some conditions and this result is supported by numerical experiments in queuing models.