Methods for System Selection Based on Sequential Mean–Variance Analysis

We propose two sequential, indifference-zone procedures for the comparison of simulated systems. Comparisons and selection of the best system are based on the mean and variance of a performance metric estimated by simulation. The mean represents the central tendency while the variance is the surrogate for the system’s inherent systematic risk. The first procedure identifies the system(s) with the largest expected value and smallest variance. The second procedure uses the variance of a reference system as a risk threshold, and selects the system with the largest mean from among those with an acceptable level of risk not above the threshold. Numerical experiments demonstrate the validity and efficacy of the proposed procedures.

The online appendix is available at https://doi.org/10.1287/ijoc.2018.0808.