Extensions of the methods of hierarchical modelling and analysis are proposed through application to data sets in four different areas of study: relationship of genetic mutations to breast and ovarian cancer; diet and the development of colorectal adenomatous polyps; exposure to low levels of ionizing radiation and cancer; and relationship between chemical exposures and cancer in humans and the same exposures and cancer in animal test species. In each case data sets are available to the investigator. The method depends on expressing the target parameters to be estimated as functions of more basic parameters so that these postulated relationships can be used to """"""""incorporate similarities"""""""" among the target parameters so as to obtain """"""""more plausible and stable estimates"""""""" of their effects. The extensions to the methods depend to some extent on the various data sets but include multistage modeling of the relationship between the target parameters and the more basic parameters, by grouping them or by adding a further level of them and by grouping the dependent binary variables where several of them are involved, e.g. different sites of disease. A variety of checks on sensitivity and performance is planned for each of the four areas of study, including checks on different choices of the parameters in the higher models and simulation of the results, using the basic data sets as guides to """"""""true"""""""" structure.
Showing the most recent 10 out of 17 publications
