Based on extended negative binomial thinning operator, an integer-valued INAR(1) model is constructed by prespecifying the distribution for the innovation process. The probabilistic properties of the model are given and the parameters of the model are estimated by using the quasi likelihood estimation method, and least squares and maximum likelihood estimation methods are also considered. The validity of these estimation methods is evaluated by numerical simulation, and the application of the model is given based on the actual data. Through comparison, it is concluded that the INAR (1) based on the extended negative binomial thinning operator with the innovation process of geometric distribution is a more suitable model for the data.
张庆春, 张黎, 范晓东.
整数值时间序列的拟似然推断
[J]. 吉林化工学院学报, 2021, 38(11): 89-93.
Zhang Qingchun, Zhang Li, Fan xiaodong.
Quasi-likelihood Inference for Integer-valued Time Series
. Journal of Jilin Institute of Chemical Technology, 2021, 38(11): 89-93.