A New Bivariate Binomial Time Series Model
2019, v.25, Issue 2
In this manuscript we introduce a new bivariate time series model whose both components have identical binomial distributions. We study some properties of the model and prove that
the model is strictly stationary and ergodic. The correlation structure of the model is presented and thereafter we study different estimation procedures such as
Yule\tire Walker, conditional least squares and conditional maximum likelihood. Simulation studies show that the best estimates in the sense of lower standard deviations are obtained by the maximum likelihood estimation method. We show that the model can be used as a good choice for modelling the number of rainfall days per week. The paper ends with a discussion of a possible extension of the introduced model.
Keywords: bivariate binomial time series; binomial marginals; asymptotic properties; stationarity; ergodicity; estimation