ABSTRACT
This chapter introduces a shifted Generalized Beta of the second kind (GB2) regression model, which is able to capture the skewness of the distribution of the fair market values. In comparison to kriging methods, the GB2 regression model does not require calculating the distances between the variable annuity policies. The chapter explains the GB2 regression model for predicting fair market values of variable annuity policies. It describes the maximum likelihood method for parameter estimation with the multi-stage optimization method. Minimizing the negative log-likelihood function is equivalent to maximizing the log-likelihood function. The validation measures indicate that results produced by the GB2 regression model with conditional Latin hypercube sampling are worse than those produced by the GB2 regression model with Latin hypercube sampling. The chapter also introduces the GB2 distribution to model the fair market values in order to capture the skewness.
