|
Activity Number:
|
410
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Wednesday, August 9, 2006 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #306630 |
|
Title:
|
Bayesian Spatial-Temporal Smoothing of Cancer Mortality Rates
|
|
Author(s):
|
Gentry White*+ and Dongchu Sun and Paul Speckman
|
|
Companies:
|
University of Missouri-Columbia and Virginia Polytechnic Institute and State University/University of Missouri-Columbia and University of Missouri-Columbia
|
|
Address:
|
146 Middlebush, Columbia, MO, 65211,
|
|
Keywords:
|
Bayesian ; thin-plate splines ; spatio-temporal ; cancer mortality rates
|
|
Abstract:
|
The high-quality data on cancer available covers a sufficient span of time for the consideration of temporal trends in incidence and mortality. Recent efforts toward more in-depth analysis of the data using spatial and temporal modeling include Clayton and Kaldor (1987)---who used a conditional auto-regressive (CAR) prior for the spatial effects---and van der Linde et al. (1995)---who used a thin-plate, spline-based prior for spatial effects. The model presented here implements a semiparametric spatio-temporal prior using a thin-plate spline prior for spatial effects and an intrinsic auto-regressive prior for temporal trends. A prior whose covariance matrix is defined as the Kronecker product of the spatial and temporal priors is introduced. Results are compared to a similar joint model, using a CAR prior for the spatial effects.
|
