|
Abstract:
|
In epidemiologic research, exposures often have several aspects among which we wish to discriminate. For instance, many environmental exposures can be separated into years of exposure, maximum intensity of exposure, cumulative exposure, and age at first exposure, to name a few. One characteristic of these aspects is they are usually correlated. This characteristic makes it difficult to evaluate the independent contributions of the individual aspects. To draw distinctions among these aspects it is necessary to control for the others, as completely as possible, while evaluating one. This often results in conflict between control of individual characteristics and interpretation of their relationship to the risk of disease. A hybrid method that provides both interpretability and discrimination is described. This method is evaluated using a simulation. As an example of its use, the relationship between aspects of cigarette smoking and lung cancer is investigated.
|
