Abstract
The article introduces a first-order bivariate integer-valued moving average process (BINMA(1)) where the respective innovation series are marginally COM-Poisson distributed under nonstationary moments. The purpose of this process is to model inter-related INMA(1) time series that are known to exhibit different levels and types of dispersion. The unknown parameters of the model are estimated by the Generalized Quasi-likelihood (GQL) approach. Simulation experiments and a real-life data application to intra-day stock series of two banking companies is presented.
Keywords: