Multiple-Predictor Smoothing Methods For Sensitivity Analysis; SANDIA REPORT, SAND2006-4693

Various paginations (approximately 130 pages). Figures. Tables. References. Format is 8.5 inches by 11 inches. The use of multiple predictor smoothing methods in sampling-based sensitivity analyses of complex models is investigated. Sensitivity analysis procedures based on smoothing methods employing the stepwise application of the following nonparametric regression techniques are described: (1) locally weighted regression, (2) additive models, (3) projection pursuit regression, and (4) recursive partitioning regression. The indicated procedures are illustrated with both simple test problems and results from a performance assessment for a radioactive waste disposal facility. Smoothing procedures based on nonparametric regression techniques can yield more informative results than can be obtained with more traditional sensitivity analysis procedures based on linear regression, rank regression or quadratic regression when nonlinear relationships between model inputs and predictions are present.…