|
Activity Number:
|
214
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2009 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Physical and Engineering Sciences
|
| Abstract - #305871 |
|
Title:
|
Quantile Estimation with Computer Experiments
|
|
Author(s):
|
Scott Vander Wiel*+ and Jim Gattiker
|
|
Companies:
|
Los Alamos National Laboratory and Los Alamos National Laboratory
|
|
Address:
|
MS F600, Los Alamos, NM, 87545,
|
|
Keywords:
|
computer model ; Gaussian process ; emulator
|
|
Abstract:
|
Physical phenomena are often modeled in complex computer codes that take a known vector of inputs and compute a corresponding output quantity. If the inputs have a known probability distribution then the output distribution could, in principle, be determined by brute force Monte Carlo simulation. Tail quantiles such as the "1 in 500" value are sometimes the focus of product design reviews and therefore methods are needed to estimate the output extremes even if the computer codes can only be run at far fewer than 500 input settings. In these cases a more efficient approach is to estimate a distribution of plausible emulator functions for the code and analyze the output distributions of the emulators.
|
