A1228
Title: A new two-component hybrid model for highly right-skewed data sets with applications.
Authors: Patrick Osatohanmwen - Free University of Bozen-Bolzano (Italy) [presenting]
Abstract: Random processes resulting in data sets with heavy tails are very common in many real life situations. When a data set has a tail that is highly skewed to the right, we can decompose the observations in the data set into two components namely: a main innovation component which contains observations of the data set around the mean and a right tail component which accounts for extreme observations in the data set. This situation makes a single univariate distribution or a non-parametric model very inefficient in fitting the data set. More so, the data set would require a distribution with two components each handling the part of the data set for which it is well suited. In this paper, we propose a new two-components distribution which links the half-Cauchy distribution (for the main innovation) and the generalized Pareto distribution (for the tail) to define a hybrid distribution which is suitable for fitting highly right skewed non-negative data sets. An unsupervised iterative estimation scheme based on the Levenberg-Marquardt algorithm is adopted for the estimation of the parameters of the hybrid distribution. An application of the new hybrid distribution in finance and hydrology is carried out.
