项目名称: 多元线性整值时间序列的统计分析
项目编号: No.11301212
项目类型: 青年科学基金项目
立项/批准年度: 2014
项目学科: 数理科学和化学
项目作者: 张海祥
作者单位: 吉林大学
项目金额: 22万元
中文摘要: 多元整值时间序列数据在现实生活中是普遍存在的,然而关于此类数据的研究结果却很少。本项目主要针对多元线性整值时间序列数据展开如下的研究工作:首先,我们基于符号稀疏算子提出新的多元整值 AR, MA以及 ARMA 模型,它们可以建模取负值以及具有负相关性的数据。此外,我们将研究多元整值自回归过程具体的边际分布。其次,对于存在缺失数据的多元整值时间序列数据, 提出新的数据填补方法,同时研究缺失情形下多元整值时间序列的经验似然推断, 拟似然推断, 假设检验等问题。最后,我们将考虑多元线性整值时间序列模型的变点检验, 质量控制, outliers 识别以及模型预测等问题,此外,我们还将从频域的角度对多元整值时间序列展开研究。
中文关键词: 整值时间序列;中介效应;高维数据;纵向数据;面板计数数据
英文摘要: Multivariate integer-value time series data are very common in practice, however, there are very few results on this kind of data. This research will pay attention to the multivariate linear integer-valued time series data. Firstly, we will propose new multivarite integer-valued AR, MA and ARMA models, which can model negative integer-valued data with negative correlation. Furthermore, we will study the marginal distribution of the multivariate autoregressive processes. Secondly, we will propose new imputation method for the multivarite integer-value time series with missing data, and study the empirical likelihood, quasi likelihood inference and hypothesis test in this situation. Lastly, we will condiser the change-point test, quality control, outliers identification and model prediction for the multivariate linear integer-valued time series model. Moreover, we will study the multivariate integer-valued time series based on frequency analysis.
英文关键词: Integer-valued time series;Mediation effects;High-dimensional data;Longitudinal data;Panel count data