Spatial regression models in epidemiological studies.

*(English)*Zbl 0965.62095
Lawson, Andrew (ed.) et al., Disease mapping and risk assessment for public health. Chichester: Wiley. 203-215 (1999).

From the introduction: We consider a truncated auto-Poisson model in the analysis of prostate cancer mortality in Valencia (Spain). We have performed the regression analysis of mortality data by considering age and nitrate contamination of drinking water as covariates. The main goal is to determine the possible influence of this contamination. In Section 15.2 we present the truncated auto-Poisson distribution jointly with alternative models to be used in a subsequent comparative study. In Section 15.3 we describe the proposed MCMC optimisation procedure in general terms. In Section 15.4 we describe the results obtained in the analysis of prostate cancer mortality in Valencia from a frequentist viewpoint. We compare maximum pseudo-likelihood with MCMC maximum likelihood, and discuss how the inclusion of spatial dependence in the model makes contamination of drinking water non-significant at the usual levels. Similar results are attained when we perform a Bayesian analysis of the same model. To compare this last result with a widely accepted standard, we have adjusted GLMM to the same data from a Bayesian perspective. We then discuss the results obtained under both models.

For the entire collection see [Zbl 0942.00010].

For the entire collection see [Zbl 0942.00010].

##### MSC:

62P10 | Applications of statistics to biology and medical sciences; meta analysis |

62M30 | Inference from spatial processes |