Simulation Validation Using a Non-Parametric Statistical Method

摘要

A primary advantage to using modeling and simulation (M&S) in a test program is it can often answer test measures that, if answered using real-world data, would require unrealistically expensive, time-consuming, or complex test events. Simulation outputs, though, are only as good as the underlying assumptions and models built into the simulation. For even moderately complex simulations, it is not easy to predict the quality of simulation results based purely on the logic that if component models, theoretical component interactions, and simulation inputs are valid, then the simulation results will be as valid as the results of a real-world test event under the same conditions. Direct comparison of simulation results to real-world test data is often conducted as part of a simulation validation effort. The analyst is faced with a difficult question: has enough data been collected under these conditions to be considered a statistically significant sample for comparison with the simulation.