Navigation:  Research interests  Current projects  Selected papers  Recent talks  Memberships  (orange background shows links on this page)

Research interests

I study model-based inference for complex systems. I work with directly with experts in the Natural Sciences who want to make better predictions, and I develop the necessary statistical methodology. Rougier (2006) describes a fairly simple inferential approach, applicable to Climate Science.

One key challenge is to devise intuitive and flexible ways to describe the ways in which scientists judge that their models are informative about the underlying system; see, for example, Goldstein and Rougier (2004, 2006b). These will also allow us to unify results from several different models of the same system.

I focus mainly on large problems, where the models can be expensive to evaluate. One approach is to use the model-evaluations to construct an emulator, which is a stochastic representation of the underlying model (which is typically deterministic); see, for example, the talk Introduction to emulators, or the short paper Inference in Ensemble Experiments. Another is to adopt a Bayes Linear approach; see, for example, Goldstein and Rougier (2006a).

Current projects

These projects have a climate focus

1. The probability of rapid climate change (PI Peter Challoner, National Oceanography Centre, Southampton). Funded under the NERC-RAPID Directed Programme.
2. RAPID modelling intercomparison project (PI Jonathan Gregory, Reading and the Met Office). Funded under the same programme.
3. PaleoQUMP: using paleodata to reduce uncertainties in climate prediction (PI Sandy Harrison, Bristol). Funded under the NERC-QUEST Directed Programme.
4. Met Office. External Expert, funded by DEFRA (TBC).

These projects are more general

5. Managing uncertainty in complex models (MUCM, PI Tony O'Hagan, Sheffield). Funded by a Basic Technologies grant from the Research Councils UK.
6. SAMSI year-long program: Development, Assessment and Utilization of Complex Computer Models (Organiser: Jim Berger, Duke University, North Carolina).
7. Uncertainty Analysis for Random Computer Models (Mentor: Michael Goldstein, Durham). Statistics Mobility Fellowship, funded by the EPSRC.

Back to the top

Selected papers