Model choice considerations and information integration using analytical hierarchy process
- Los Alamos National Laboratory
- BOOKER SCIENTIFIC
- UNM
Using the theory of information-gap for decision-making under severe uncertainty, it has been shown that model output compared to experimental data contains irrevocable trade-offs between fidelity-to-data, robustness-to-uncertainty and confidence-in-prediction. We illustrate a strategy for information integration by gathering and aggregating all available data, knowledge, theory, experience, similar applications. Such integration of information becomes important when the physics is difficult to model, when observational data are sparse or difficult to measure, or both. To aggregate the available information, we take an inference perspective. Models are not rejected, nor wasted, but can be integrated into a final result. We show an example of information integration using Saaty's Analytic Hierarchy Process (AHP), integrating theory, simulation output and experimental data. We used expert elicitation to determine weights for two models and two experimental data sets, by forming pair-wise comparisons between model output and experimental data. In this way we transform epistemic and/or statistical strength from one field of study into another branch of physical application. The price to pay for utilizing all available knowledge is that inferences drawn for the integrated information must be accounted for and the costs can be considerable. Focusing on inferences and inference uncertainty (IU) is one way to understand complex information.
- Research Organization:
- Los Alamos National Laboratory (LANL), Los Alamos, NM (United States)
- Sponsoring Organization:
- USDOE
- DOE Contract Number:
- AC52-06NA25396
- OSTI ID:
- 1038876
- Report Number(s):
- LA-UR-10-07004; LA-UR-10-7004; TRN: US201209%%28
- Resource Relation:
- Journal Volume: 2; Journal Issue: 6; Conference: 2010 Sensitivity Analysi of Model Putput (SAMO) ; July 10, 2010 ; Milam, Italy
- Country of Publication:
- United States
- Language:
- English
Similar Records
Prioritizing Project Risks Using AHP
A flexible numerical approach for quantification of epistemic uncertainty