Statistics Seminars: Policy making using computer simulators for complex physical systems; Bayesian decision support for the development of adaptive strategies
14 December 2009 14:15 in CM221
Policy problems, such as those involving mitigation for climate change, often require the use of computer simulators. These computer models mimic complex physical systems and are used to provide insight into how the system will behave under any given policy. We consider the problem where a policy is to be made and where, at a number of fixed points in the future, the system may be observed and the policy may be adapted accordingly. This problem defines a decision tree with infinitely many branches and where the computer simulators are informative for the probabilities on any given branch. We describe a method we call Sequential Emulation that leads to powerful decision support for policy problems. Our methods allow future observations of the complex system along with runs on future improved models for the system to influence the support we offer today.
Contact email@example.com for more information