Some gamma processes based on the dirichlet-gamma transformation

Abstract
The purpose of this paper is to develop some extensions of the beta-gamma processes introduced by Lewis et al. (1989). We consider a univariate gamma ARMA(p,p-l) process and multiple gamma AR(1), MA(1) and ARMA(1,1) processes. These models are useful for modelling or generation of sequences of dependent (multiple) gamma random variables which arise in various fields. In developing each of these models we make use of the Dirichlet-gamma transformation. Several properties of the models such as autocorrelation, joint distribution, regression and time reversibility are investigated. The analogy between the gamma and the negative binomial distributions makes it possible to develop similar models for the negative binomial distribution.

This publication has 7 references indexed in Scilit: