Binary generalized extreme value additive modelling

Logistic regression is the commonly used model for bankruptcy prediction of small and medium enterprises, for instance. However, the assumptions of symmetric link function and linear or pre-specified covariate-response relationships may not be realistic, especially in scoring applications. To deal with these issues a binary generalized extreme value additive model is introduced. The approach uses the quantile function of the generalized extreme value distribution as link function as well as smooth functions of continuous predictors to flexibly model their effects. The framework is implemented in the bgeva R package which has a bgeva() function that works in a similar way to the glm() and gam()-like functions in R. The main ideas behind the methodology will be discussed and the bgeva package illustrated using Italian data on small and medium enterprises.