3 Propagation of Uncertainty

Propagation of uncertainty is a fundamental concept in several aspects related to modeling.

• Uncertainty in data values (often termed experimental error) leads to uncertainty in model coefficient values and uncertainty in model coefficient values leads to uncertainty of the in-use model-calculated result.

• Prior to regression, in design of experiments propagation of uncertainty is applied to experimental measurements to ensure adequate precision in the experimental design, to define the number of data sets required for adequate precision, and to determine experimental conditions that need additional focus.

• In regression, propagation of uncertainty can be used to select regression convergence criteria.

• In model development, uncertainty analysis is useful in understanding the impact of linearization, in assessing model utility, and in relating data uncertainty to model coefficient uncertainty.

• In model validation, uncertainty analysis is used to compare modeled residuals to expected trends and magnitudes.

• In model use, propagation and reporting uncertainty on model-calculated results is important to the user, who must accommodate uncertainty of calculated values to make decisions.