Gosling, J. P. and O'Hagan, A.
University of Sheffield
Publication details: Not published. This is a supporting document for Kennedy et al (2006).
Uncertainty analysis is the evaluation of the distribution of a computer code output Y given uncertainty about the code’s true inputs X. When we sum the outputs of a number of different codes, it is important to appreciate the different sources of uncertainty that can arise. An estimate of the biospheric carbon flux for England and Wales re- quires the outputs of many different computer codes to be aggregated. In this report, the techniques needed for output aggregation will be discussed and the process of producing estimates of the carbon flux will be reviewed. This includes an independent normal approximation to a Dirichlet distri- bution for proportion parameters.