We investigate a new bivariate-integer valued moving average time series process where the innovation series follow the bivariate Poisson assumption under stationary moments and constant cross-correlations. Furthermore, due to the complication involved in specifying the joint likelihood function, this paper considers a robust generalized quasi-likelihood approach to estimate the mean, serial and dependence parameters. Unlike previous estimation techniques such as the Generalized Least Squares, this estimation approach here involves a two-step Newton-Raphson iterative procedure where in the first step, the serial and cross correlations are estimated while in the second step, these dependence estimates are used to compute iteratively the vector of regression coefficients. The consistency of the estimates under this approach is checked through several simulation experiments under different combinations of low and high serial and cross-correlations.

中文翻译：

---

使用鲁棒的GQL方法分析完整的BINMA时间序列过程

我们研究了一个新的二元整数值移动平均时间序列过程，该过程的创新序列在平稳矩和恒定互相关下遵循二元泊松假设。此外，由于指定联合似然函数所涉及的复杂性，本文考虑了一种鲁棒的广义拟似然法来估计均值，序列和依赖参数。与以前的估计技术（例如，广义最小二乘）不同，此估计方法涉及两步牛顿-拉夫森迭代过程，其中第一步是估计串行和互相关，而在第二步中，这些相关性估计用于迭代计算回归系数的向量。