Stefano Conti, John Paul Gosling, Jeremy Oakley and Anthony O'Hagan
Centre for Health Economics, University of York, York, England and Department of Probability and Statistics, University of Sheffield, Sheffield, England
Publication details: Submitted to Biomatrika, 2007.
Computer codes are used widely in scientific research to study and predict the behaviour of complex systems. The run times of computationally- intensive computer codes are often such that they are impractical to run the thousands of times as is conventionally required for sensitivity analysis, uncertainty analysis or calibration. In response to this problem, efficient techniques have been developed based on a statistical representation of the computer code. The approach, however, is less straightforward for dynamic computer codes, which are designed to represent time-evolving systems. We develop an iterative system to build a statistical model of dynamic computer codes, which is illustrated with an application to a rainfall-runoff simulator.
Keywords: Bayesian inference; Computer experiments; Dynamic simula- tors; Emulation; Gaussian process; Recursive modelling