Joint Research Conference
June 24-26, 2014
Data collected from correlated processes arise in many diverse application areas including both computer and physical experiments, and studies in environmental and ecological science. Often, such data are used for predicting the process at unobserved points in some continuous region of interest, for example, constructing an emulator of a computationally expensive simulator. The design of the experiment from which the data are collected may strongly influence the quality of the model fit and hence the precision of subsequent predictions. The objective of this work is to obtain an optimal choice of design to allow precise prediction of the response at unobserved points using an appropriate statistical model.
We consider Gaussian process models that are typically defined by a spatial correlation structure that may depend upon unknown parameters. This parametric uncertainty may affect the choice of design points, and ideally should be taken into account when choosing a design.
We illustrate a new approach for the selection of Bayesian optimal designs using a decision theoretic approach and an appropriate loss function. To avoid the computational burden usually associated with Bayesian design, we have developed and investigated a new, computationally inexpensive, closed form approximation of the objective function. The resulting designs are illustrated through a number of examples using applications from both computer experiments and spatial statistics.
Joint work with Maria Adamou, Sue Lewis and Sujit Sahu (University of Southampton)
ype your paragraph here.