Modelling stock data using a geometric BINAR(1) model

The intricate cross-correlation between the counting series makes the non-stationary bivariate integer-valued autoregressive of order 1 (BINAR(1)) model with a full correlation structure very challenging to model. This paper offers a novel full correlation nonstationary BINAR(1) model with geometric marginals (BINAR(1)GEOM). The regression and dependency parameters are estimated using the Generalized Quasi-Likelihood (GQL) estimation approach, which is contrasted with several well-known estimation techniques: the Generalized Method of Moments (GMM) and Generalized Least Squares (GLS). A numerical experiment is conducted to compare the performance of these estimation methods and the proposed model is tested on some stock transactions in the Mauritian market.