A case study in hierarchical modeling: Agriculture and prostate cancer in Iowa

Epidemiologists who study cancer etiology are often asked to conduct etiologic investigations of environmental agents and cancer in populations. The difficulty in estimating the effects of relevant exposures arises because environmental data are usually available only on an aggregate level, such as by county. Individual exposures to chemical agents are usually unknown, as is the joint distribution between the environmental exposure level and disease status. Frequently, some information on the cancer cases themselves is available from a disease registry which typically includes age at diagnosis, gender, and race, and these factors may require control in the analysis. An appropriate statistical model which incorporates this nested data structure is needed. This paper illustrates application of a special case of mixed model, namely hierarchical model. to the study of agricultural factors and prostate cancer in Iowa.