A comparison of sensitivity analysis techniques.

Abstract

Modeling the movement and consequence of radioactive pollutants is critical for environmental protection and control of nuclear facilities. Sensitivity analysis is an integral part of model development and involves analytical examination of input parameters to aid in model validation and provide guidance for future research. Sensitivities of 21 input parameters have been analyzed for a specific-activity tritium dose model using fourteen methods of parameter sensitivity analysis. This report demonstrates, for each sensitivity method, the required calculational effort, the sensitivity ranking of parameters, and the relative method performance. The sensitivity measures include the following: partial derivatives, variation of inputs by 1 standard deviation (SD) and by 20%, a sensitivity index, an importance index, a relative deviation of the output distribution, a relative deviation ratio, partial rank correlation coefficients, standardized regression coefficients, rank regression coefficients, the Smirnov test, the Cramer-von Mises test, the Mann-Whitney test, and the squared-ranks test.