Fall 2011 Seminar Series

DATE: September 29th
LOCATION: E62-550
TIME: 4:15pm
Reception immediately following

SPEAKER:
Barry Nelson

TITLE
Quantifying Input Uncertainty in Stochastic Simulation

ABSTRACT
Stochastic simulations consist of input models and logic. The input models are often probability distributions fit to real-world data, in which case the simulation results are conditional on the chosen distributions. Standard practice in simulation output analysis neither quantifies nor even acknowledges that uncertainty about the input models may dominate everything else. We present a framework for producing confidence intervals that account for uncertainty about the input models. To achieve this goal we introduce metamodel-assisted bootstrapping, and illustrate its performance relative to other proposals for dealing with input uncertainty.