Statistical modeling of epidemiologic data

Abstract

The application of statistical modeling to epidemiology may help suggest a form for the mechanism of exposure action. But distinguishing between the entertained biological models is often difficult due to both inadequacies in epidemiologic studies and inaccuracies in the verbal specifications of the hypothesized interaction mechanisms. For example, the independent and interactive effects of asbestos and smoking on the production of lung cancer have not yet been fully established. In the present communication an analysis of illustrative data from a hypothetical case-compeer study was attempted with the estimation of rate ratios and the use of a log-linear model fitting technique. These analyses allow a parametric representation of the testable models. For adequate material they might provide tentative insight as to whether the data would conform more closely to an additive model than to a multiplicative one or to some other advocated pattern of action.