Use of a Single Regression Model to Estimate Missile System Development Cost

Abstract Current cost estimating methods rely heavily on the use of regression models, especially for programs in the Development phase of the life cycle. Estimating costs using regression models is preferred during the early life cycle of a program when much is unknown about the program being estimated. Costs related to unknown-unknowns, those costs whose source and magnitude cannot be described prior to their occurrence, must be captured if an estimate is to be accurate. Because regression models utilize historical costs for similar cost elements being estimated, they inherently include many of these unknowns. Thus, perhaps the best way to minimize uncaptured cost is to utilize regression models which estimate at the highest feasible level of the cost estimating structure, thereby reducing the potential for unknown-unknowns. This paper demonstrates this approach through the derivation of a single sound and validated regression model for use in estimating the development cost of ballistic missiles.