Activity Number:
|
354
|
Type:
|
Contributed
|
Date/Time:
|
Wednesday, August 14, 2002 : 2:00 PM to 3:50 PM
|
Sponsor:
|
Section on Bayesian Stat. Sciences*
|
Abstract - #301417 |
Title:
|
Treatment of Edge Effects in Spatial Data Analysis via Bayesian Markov Field Process Convolution Models
|
Author(s):
|
John Kern II*+ and David Higdon
|
Affiliation(s):
|
Duquesne University and Los Alamos National Laboratory
|
Address:
|
600 Forbes Avenue, Pittsburgh , Pennsylvania, 15282,
|
Keywords:
|
Bayesian ; Markov random field ; process convolution
|
Abstract:
|
Spatial moving average models offer researchers an appealing option when modeling spatial processes; recent research has documented their ability to describe both stationary and non-stationary processes, as well as their flexibility in modeling a wide class of variograms. Our research examines the impact of edge effects in spatial analyses through the use of spatial moving average models. In particular, we model the spatial process of interest as the convolution of a discrete Markov random field (defined only over the domain of the spatial process) with a kernel that is allowed to vary according to the distances between the kernel's center and the edges of the spatial domain. This approach is shown to improve the process predictions along the boundaries of the spatial domain.
|