The statistical uncertainty resulting from the lack of knowledge of which model represents a given stochastic process is analyzed. This analysis of model uncertainty leads to a composite Bayesian distribution. The composite Bayesian distribution is a linear model of the individual Bayesian probability distributions of the individual models, weighted by the posterior probability that a particular model is the true model. The composite Bayesian probability model accounts for all sources of statistical uncertainty, both parameter uncertainty and model uncertainty. This model is the one that should be used in applied problems of decision analysis, for it best represents the knowledge, or lack of it, to the decision maker about future events of the process.
All Science Journal Classification (ASJC) codes
- Water Science and Technology