| Denison DG, Holmes CC. Bayesian partitioning for estimating disease risk. Biometrics 2001; 57: 143-149. |
....on posterior inference. Under Bayesian model averaging, the posterior mean risk surface from any partition model can provide a smooth estimate of the risk surface. Models that can be described in this framework include the clustering or segmentation models of Knorr Held and Ra er (2000) and Denison and Holmes (2001), which propose di erent spatial models for fz i g. In the model 4 investigated in this paper we propose to use a Potts model for fz i g, with the number of states and strength of interaction unknown. In contrast to the partition models cited, we retain a Markovian structure for the fz i g. Yet ....
Denison, D. G. T., and Holmes, C. C. (2001) Bayesian partitioning for estimating disease risk. Biometrics, 57, 143-149.
....region is assigned to, is minimal a priori. This indicates that, in contrast to the Markov random eld models, there is no strong dependence of the prior amount of smoothing on the number of neighbours. Related formulations based on continuous Voronoi tesselations have recently been proposed in Denison and Holmes (1999) for disease mapping. If applied to the usual areal data, their formulation has the disadvantage that clusters will not necessarily be connected. The discrete Knorr Held and Ra er (2000) model fully acknowledges the discrete nature of space and automatically ensures that all clusters are ....
Denison, D.G. and Holmes, C.C. (1999). Bayesian partitioning for estimating disease risk. Technical Report, Imperial College London.
No context found.
Denison DG, Holmes CC. Bayesian partitioning for estimating disease risk. Biometrics 2001; 57: 143-149.
Online articles have much greater impact More about CiteSeer.IST Add search form to your site Submit documents Feedback
CiteSeer.IST - Copyright Penn State and NEC