A predictive random effects model of dependent claims frequency and severity
Himchan Jeong (University of Connecticut)
Friday, February 9, 2018
In constructing our predictive random effects model with dependence, we focus on families of distributions with conjugate random effects in order to be able to derive explicit moment formulas for the aggregate claims. In our search for such models for claims severity, we discovered that the family of GB2 distributions could be derived with an underlying distribution based on generalized gamma (G-Gamma) and a random effect based on generalized inverse gamma (GI-Gamma). This resulting family provides a flexible parametric distribution with four parameters describing scale and various shapes.