Year: 2023
Type: conference-paper
Abstract: Quantify variance in model output and attribute it to different uncertainty sources Variance-based global sensitivity analysis Identical inputs produce different results Stochastic computational modelsSeparate variance contributions from stochastic solver and uncertainty sources Variance deconvolution OVERVIEW 2/19 BACKGROUND -GLOBAL SENSITIVITY ANALYSIS 3/19 β’ Consider π = π π !, π " , π # , function of 3 uncertainty sources β’ Statistics with respect to π β’ πΌ $ π , mean β’ πππ $ π , variance BACKGROUND -GLOBAL SENSITIVITY ANALYSIS 3/19 β’ Consider π = π π !, π " , π # , function of 3 uncertainty sources β’ Fix π ! as constant β conditional mean πΌ $ !,$ " π|π != ππππ π‘πππ‘ β’ Statistics with respect to π β’ πΌ $ π , mean β’ πππ $ π , variance Conditional statistics for π !
DOI: 10.2172/2430477
Authors: Kayla Clements, Gianluca Geraci, Aaron Olson, Todd Palmer
Institutions: Sandia National Laboratories, Oregon State University
Language: English
Subfield: Statistics, Probability and Uncertainty
Field: Decision Sciences
Domain: Social Sciences
Open Access status: gold
OpenAlex ID: W4392737248