Misturas finitas de distribuições hiperbólica generalizada normal assimétrica

Abstract

In this work, a flexible model of finite density mixtures is presented, based on the class of asymmetric generalized hyperbolic distributions, in which some important properties and results are determined, so that an estimation study by maximum likelihood of the parameters is subsequently carried out. present in the proposed model. For this, the EM algorithm was used, obtaining the estimates of the model parameters in an iterative way. In addition, some particular cases are discussed, including additional comments on the implementation of this algorithm. A simulation study is also designed to evaluate its performance, as well as four applications to real data, which illustrate the behavior of the model. The results obtained showed a strong possibility in the use of this model in statistical analyzes involving data that present significant multimodality and non-Gaussian behaviors, such as asymmetry and heavier tails.