Dynamic multistage software estimation

This paper develops dynamic models that advance the state-of-the-art in software estimation to meet the analytic requirements of the most current spiral and prototyping software development process paradigms. These models account for the dynamics of changing requirements, system design, and other policy factors. The dynamic formulation permits analysis of the effect of current-stage decisions on future decision opportunities in light of multiple objectives associated with cost overrun and schedule delay. A linear-normal dynamic model with closed-form solution is developed first and provides the context for defining the components of the software cost-estimation dynamic model and for describing the interaction of these components. A second model relaxes the linearity and normal distribution restrictions, employing nonlinear state and observation equations derived from the Intermediate COCOMO and COCOMO II models.