Modelling Educational Data Series using a Novel High-Ordered BINAR Model

Abstract

This paper treats the modelling of a bivariate integer-valued autoregressive of order p (BINAR(p)) model under non-stationary moment conditions where the inter-relation between the series is induced by pair of related Poisson innovations. This novel model is more convenient to model time series data on Education. The model parameters constituting of the mean or regression effects and the correlation effects are estimated via a generalized quasi-likelihood (GQL) equation since the joint and marginal distribution of the counting series under the non-stationary conditions is difficult to specify. The proposed model is used to analyse a series of educational data.