|
Activity Number:
|
308
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 14, 2002 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistics & the Environment*
|
| Abstract - #301722 |
|
Title:
|
Optimal Bayesian Network Designs in Spatial Statistics
|
|
Author(s):
|
Milena Banjevic*+ and Paul Switzer+
|
|
Affiliation(s):
|
Stanford University and Stanford University
|
|
Address:
|
POB 14440, Stanford, California, 94309, USA Sequoia 136, Stanford, California, 94305, USA
|
|
Keywords:
|
spatial statistics ; network design ; bayesian design ; optimal sampling
|
|
Abstract:
|
We consider the problem of designing a network of sampling locations in a spatial domain that will be used to interpolate a spatial field. The optimal design depends largely upon the parameters of the random field model, which are rarely explicitly known. We express this uncertainty through an appropriate class of prior distributions. We then introduce a Bayesian sequential sampling algorithm. Optimal locations are selected sequentially, using a minimum estimation variance criterion. At each step, posterior parameter values are updated through realizations from previously selected locations. We focus on the random field model in which variance is given by an unknown step function of the locations. This model is appropriate for some environmental applications, such as measurements taken over land/sea, wood/rock, or urban/rural areas. We examine the convergence of parameter estimates for different prior distributions and cardinalities of the design network. We discuss the improvement of the Bayesian method over conventional sampling techniques, which assume known parameter values. Simulated examples and an application to temperature variation in the Mojave desert are also presented.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2002 program |
