Time series models for non-Gaussian processes
H. W. Block, Naftali A. Langberg, David S. Stoffer · Lecture notes-monograph series · 1990
In this paper we present univariate and multivariate time series models for processes with non-Gaussian marginal distributions.These include bivariate autoregressive-type models for processes with bivariate exponential marginals, nonlinear autoregressive-type models for processes with Dirichlet marginals, and nonlinear models for univariate time series with arbitrary marginal distributions.Examples of applications to real data sets are given for some of the models discussed.When applicable, the theory of positive dependence is used to establish the association of the processes.