Composite models with underlying folded distributions

Abstract In this note, we examine the performance of 25 new composite models that are derived from 5 underlying folded distributions for modeling insurance loss data. These models are assessed using standard selection criteria involving the Akaike Information Criteria and the Bayesian Information Criteria as well as proximity to empirical risk estimates. Three models are found significant in improving the goodness-of-fit than the latest development in the literature with two models reliable for risk estimation.