项目名称: 整数值时间序列数据的建模方法研究
项目编号: No.11271155
项目类型: 面上项目
立项/批准年度: 2013
项目学科: 数理科学和化学
项目作者: 王德辉
作者单位: 吉林大学
项目金额: 67万元
中文摘要: 整数值时间序列在生产、实践中广泛存在(例如医院住院的病人数、一组保单的理赔次数等),对其统计分析是国际上研究的一个热点问题.由于数据本身的特征(整数值、相关性),其统计分析研究还处于起步阶段,本项目旨在从数据自身特征出发,构造新模型并讨论参数的估计方法及估计量的渐近分布,进而构造检验统计量,对模型的平稳性进行检验。另外,基于thinning算子的模型是处理整数值时间序列数据的基本模型,但是其在刻画形形色色的整数值数据的内在特征时存在不足,为此,我们将构造新的整数值时间序列模型(如MA、ARMA、线性GARCH、非线性GARCH 等模型),并考虑模型的建模、假设检验及模型选择等研究问题,将模型的统计推断结果应用于生产实践中,为实际工作者提供理论指导。
中文关键词: 整数值;时间序列数据;建模方法;;
英文摘要: Integer-valued time series data are commonly observed in real-world applications, whose statistical analysis has been a hot topic in the last two decades. Because of the characteristics of data (integer-valued, correlation), there is only a few research related to this topic. This projec t will aim to model these inherent features, construct new models and discuss estimators of corresponding parameters and their asymptotic distributions, an d furthermore construct test statistics to test stationarity. In addition, thinning operator models are the basic ones to deal with integer-valued time series data, but they have shortcomings and limitations in modeling various character istics of integer-valued data. We will construct new integer-valued time series models (such as MA、ARMA、linear GARCH and nonlinear GARCH models) and consider inference problems, including modeling, testing and model selection. These new resu lts will be applied to real-life practices, and provide theoretic guide for practitioners.
英文关键词: integer-valued;time series data;modeling methods;;