Statistical Challenges to Linking the Spatial Pattern of Cancer to Radiation Exposure

• Activity Number: 564
• Type: Topic Contributed
• Date/Time: Wednesday, August 1, 2012 : 2:00 PM to 3:45 PM
• Sponsor: International Indian Statistical Association
• Author(s): Stephen Rathbun*+ and Sara Wagner and John Vena
• Companies: University of Georgia and University of Georgia and University of Georgia
• Keywords: Hierarchical spatial model ; Measurement error ; Left-censoring ; MCMC algorithm ; Bayesian inference

Abstract

The epidemiological investigation of the link between cancer and exposure to radon gas in homes and uranium in drinking water poses a number of challenges to spatial data analysis. Radon and uranium levels are left-censored as some observations fall below the minimum detection limits of the instruments used to assay their concentrations. Moreover, cancer case, radon and uranium data were spatially misaligned in the sense that each sample site contains only data from one of the three variables. Therefore, a model for the effects of the two radionuclides must rely on spatial prediction of exposure levels at the locations of cancer cases. Errors in exposure estimates can lead to biased estimates of the effects of radon and uranium, and low statistical power for detecting those effects. We propose a hierarchical spatial model for the relationship between cancer cases and exposure to household radon and well-water uranium radiation that takes into account the left-censoring of the radionuclide data and spatial misalignment while correcting for measurement errors attributed to the use of predicted exposure levels.