Discrete Optimization of Statistical Sample Sizes in Simulation by Using the Hierarchical Bootstrap Method
This work addresses variance reduction in statistical simulations for researchers using hierarchical bootstrap methods, but it appears incremental as it applies dynamic programming to an existing technique.
The paper tackles the problem of optimizing sample sizes in hierarchical bootstrap simulations to reduce variance in system characteristic estimators, achieving a decrease in variance through dynamic programming.
The Bootstrap method application in simulation supposes that value of random variables are not generated during the simulation process but extracted from available sample populations. In the case of Hierarchical Bootstrap the function of interest is calculated recurrently using the calculation tree. In the present paper we consider the optimization of sample sizes in each vertex of the calculation tree. The dynamic programming method is used for this aim. Proposed method allows to decrease a variance of system characteristic estimators.