Application of Re-Weighting Techniques to Expedite Sensitivity Analysis in Monte Carlo Simulation Models. G. M. Paoli, Decisionalysis Risk Consultants, Canada; and K. P. Brand, University of Ottawa, Canada
Quantitative uncertainty analysis of even moderately complex models can pose a significant computational burden. As an example, computation times of several hours are not uncommon in current farm-to-fork microbial risk assessments. This burden tends to discourage thorough exploration of the sensitivity of model outputs to alternate distributional assumptions. In this paper we demonstrate the power of a standard re-weighting technique, which can greatly expedite such sensitivity analysis. The technique, which represents a natural extension of standard Monte Carlo methods, offers several additional advantages. Not only can analysts explore alternative distributional shapes and parameters, but a remote user of the model can explore the impact of modifying distributional assumptions without requiring re-execution of the model. This has implications for the dissemination of risk-based decision support materials as well as in improving the transparency of models through review. Further, the re-weighting approach is compatible with importance sampling (e.g., to oversample the tails of important input distributions) and can facilitate modular development and simulation strategies for large and complex models.
Go to . . .
2002 SRA Annual Meeting Table of Contents
2002 SRA Annual Meeting Author Index
Main Abstracts Menu Page
RiskWorld Home Page