Title: Parametric and Nonparametric Uncertainty Quantification

Speaker: Alyson Wilson and Dan Nordman, Iowa State University 

Date/Time: Monday, February 16, 2009 at 9am-10am 

Location: CSRI Building, Room 90 (Sandia NM) video conf to CA (915/S145 contact Hailey Poole)

Brief Abstract: This talk will consider two different approaches to uncertainty quantification. The first approach arose in response to the Science-Based Stockpile Stewardship—or more specifically, the problem of how to assess the reliability of the physics package of nuclear weapons without full-system testing. The systems in the nuclear stockpile are complex and dynamically aging, which leads to problems in assessing system reliability over time using information from a variety of sources. This modeling is heavily parametric, and focuses on how to integrate multiple sources of information (computer simulations, historical data, surrogate data, expert judgment, etc.). The approach has been applied not just to nuclear weapons, but to a variety of complex engineered systems. The second approach describes some problems with quantifying uncertainty in data exhibiting dependence. Specifically, data collected overtime and/or at different locations are often correlated, and the relationship between samples is often complex. Many statistical approaches require specifying a probability model to explain how data were generated, but such model building can be difficult with dependent data and an additional danger is that inference drawn from a misspecified model may be unreliable. The presented methods and results target quantifying uncertainty in inference without assumptions on the data dependence structure or data-generating mechanism.

Alyson Wilson is an Associate Professor in the Department of Statistics at Iowa State University and a Scientist 5 in the Statistical Sciences Group at Los Alamos National Laboratory. http://www.stat.iastate.edu/directory/personal.php?id=agw

CSRI POC: Scott Mitchell, (505) 845-7594