|
Abstract:
|
Many problems in the earth, energy, and environmental sciences involve the analysis of large quantities of data, where there is a necessary interest in each individual element. The data in these problems are often subject to various types of structure and, in particular, strong dependence. However, there is a need for inferential methods in this area in order to quantitatively interpret results over regions for scientists and policymakers. Typical multiple comparison procedures often require a compromise on the structure of the data, and/or on an accepted error rate. Thus, the primary goal is joint inference, where 'joint' is taken to be in the spatial as well as variable context. Placing the data in a (hierarchical) model with a Gaussian Markov random field component, we use the structural characteristics of the data to derive joint inference.
|
