A Multiple logistic Methodology for the Estimation of Risk Classification Models

In this chapter a methodology is developed to measure the quantitative relationship between the probability that a policy will terminate by death in the ensuing year and the policyholder’s characteristics at the time of application for life insurance. Due to the dichotomous nature of the response variable (death or survival for the following year). multiple regression analysis is inappropriate.1 Consequently, this study utilizes a maximum likelihood algorithm to estimate the coefficients in a multiple logistic model. The Newton-Raphson technique provides for an iterative solution of the estimated betas after initial estimates are calculated by the application of Fisher’s linear discriminant function.