Robust Modeling Based on L2E Applied to Combat Simulation Data
David Kim, (United States Military Academy), firstname.lastname@example.org
Parametric modeling based on a minimum distance criterion gives us a robust (i.e. resistant to outliers and other types of data contamination) way of analyzing the data. The integrated squared error (L2E) of David Scott is one such criterion, and a system of fitting a model to the data and assessing the model can be built upon it. Combat simulation data containing numerous variables may benefit from such a robust method of analysis since it is quite likely that some of the assumptions for building, for example, a least squares linear model may be violated in the data.
We will outline the development of such a system and demonstrate the new methodologies via the analysis of simulation results from One Semi Automated Forces (OneSAF). Specifically, the variable selection methods based on L2E are used in the analysis to provide models identifying key battle parameters for a given engagement.