INARS(p)模型的拟似然统计推断

Quasi-Likelihood Statistical Inference for the INARS(p) Model

用拟似然方法对p阶基于符号伯努利稀疏算子的整值时间序列模型参数进行估计,得出了参数修正的拟似然估计因子以及该估计因子的极限分布(可以用此极限分布对模型参数进行假设检验等统计分析),并通过数值模拟,将修正的拟似然估计与条件最小二乘估计进行了比较,结果表明,修正的拟似然估计在一定条件下明显优于条件最小二乘估计.

The authors used quasi-likelihood method to estimate the parameter of integer-valued AR （p） process with signed binomial thinning to obtain the modified-quasi-likelihood estimator of the model parame- ters, and the asymptotic distribution of this estimator, thus, we can have some statistical analysis results about the model parameter, such as hypothesis testing. And we compared the modified-quasi-likelihood estimator with the conditional least squares estimator via simulation. The results show that under some conditions, the modified-quasi-likelihood estimator is better than conditional least squares estimator.