Many inputs for simulation optimization models are assumed to come from known distributions. When such distributions are obtained from small sample sizes, the parameters of these distributions may be associated with an “uncertainty set” or ranges. The presence of this uncertainty means that one or more solutions may be optimal depending on which parameters from the set are used. In this paper, we present a graphical methodology that combines bootstrap sampling and cross-evaluation techniques to visualize the data driven support for alternative solutions for problems in which distribution parameters are estimated using small sample sizes. We illustrate the methodology using a voting machine allocation problem.
展开▼
