A Comparative Study of Determining Risk Factors

Abstract

The intention of the present talk is to contribute, by some further points, to the experiences already collected as regards the risk estimation by logistic functions. This is a widely used model in epidemiology. One of the main objects of the epidemiology is to predict for each individual the possibility of a given disease in a large population. For this purpose a set of data is collected from every person. These data - the so called risk factors - are not directly connected with the disease in question; moreover the most difficult problem is to select relevant data. From the mathematical point of view the above prediction problem is closely related to the statistical discrimination, the basic problem of which can be summarized as follows: there are several populations A₁,…, A_n and a sample of individuals from each. Upon measurements from these individuals a certain rule should then be set up, which enable us to put a new individual by its measurement values into the population to which it belongs.