Copulae and Dependence Modelling: Indirect Estimates of a Person's Risky Behavior Intensit

Abstract

In conditions of limited resources, the most affordable way to obtain information about the behavior of an individual is an interview. In this case, data on the individual's recent behavior are less susceptible to various types of bias. Mathematical models of behavior are used to estimate the intensity of behavior when only the data on the last three episodes are available. In the paper we consider the gamma Poisson model of behavior. The dependence of two intervals between successive episodes in this model is described in terms of copulae. Estimation of the copula parameter in this case directly leads to an estimate of the parameters of the intensity distribution in the population. In addition, knowledge of the copula type allows one to reveal some characteristics of the behavior described by the gamma Poisson model: episodes of such behavior are clustered. The possibilities of the proposed approach are demonstrated on model data.