{[ promptMessage ]}

Bookmark it

{[ promptMessage ]}

NASA-Systems Engineering

# Even in a well behaved statistical regression

This preview shows page 1. Sign up to view the full content.

This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: ility not possessed by SIMSYLS. For more detail, see DeJulio, E., SIMSYLS User's Guide, Boeing Aerospace Operations, February 1990, and Kline, Robert, et al., The M-SPARE Model, LMI, NS901R1, March 1990. cies. (See Section 6.5 for more on Integrated Logistics Support.) 5.4 Probabilistic Treatment of Cost and Effectiveness A probabilistic treatment of cost and effectiveness is needed when point estimates for these outcome variables do not "tell the whole story"—that is, when information about the variability in a system's projected cost and effectiveness is relevant to making the right choices about that system. When these uncertainties have the potential to drive a decision, the systems or program analyst must do more than just acknowledge that they exist. Some useful techniques for modeling the effects of uncertainty are described below in Section 5.4.2. These techniques can be applied to both cost models and effectiveness models, though the majority of examples given are for cost models. NASA Systems Engineering Handbook Systems Analysis and Modeling Issues 5.4.1 Sources of Uncertainty in Models There are a number a sources of uncertainty in the kinds of models used in systems analysis. Briefly, these are: • • • Uncertainty about the correctness of the model's structural equations, in particular whether the functional form chosen by the modeler is the best representation of the relationship between an equation's inputs and output Uncertainty in model parameters, which are, in a very real sense, also chosen by the modeler; this uncertainty is evident for model coefficients derived from statistical regression, but even known physical constants are subject to some uncertainty due to experimental or measurement error Uncertainty in the true value of model inputs (e.g., estimated weight or thermal properties) that describe a new system. As an example, consider a cost model consisting of one or more statistical CERs. In the early phases of the project life cycle (Ph...
View Full Document

{[ snackBarMessage ]}

Ask a homework question - tutors are online