Simulation validation with historic outcomes

Combat, unlike many real-world processes, tends to be singular in nature. That is, there are not multiple occurrences from which to hypothesize a probability distribution model of the real-world system. Mission-level models may offer more flexibility on some measures due to their extended time frame. Additionally, the parameters involved in the mission-level model may be unchanged for significant stretches of the total simulation time. In these cases, time periods may be devised so that the periods hold sufficiently similar traits such that the incremental results may be assumed to come from a common distribution. This paper details a new statistical methodology for use in validating an agent-based mission-level model. The test is developed within the context of the Bay of Biscay agent-based simulation and uses the monthly data from the extended campaign as a basis of comparison to the simulation output.