项目名称: 复杂数据下半参数双重回归模型的统计推断及其应用
项目编号: No.11301485
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: 徐登可
作者单位: 浙江农林大学
项目金额: 22万元
中文摘要: 随着人们对现实世界深入的认识,我们研究的实际数据也越来越复杂,如果只用简单的统计模型来描述已不能满足实际要求. 因此我们很有必要针对这些复杂现象,采用比较复杂全面的模型来描述.半参数双重回归模型就是其中的一类,这个模型主要的特点就是体现在对方差的重视.当前文献大多数集中于均值回归模型的统计推断,对方差参数(或者散度参数)赋予一个模型结构后,有关双重回归模型的变量选择、贝叶斯估计等统计推断却鲜有研究,但近年来也引起了很多学者的重视.因此本项目拟研究复杂数据下半参数双重回归模型的变量选择、贝叶斯估计等统计推断问题.重点讨论在复杂数据下半参数双重回归模型有关变量选择的理论性质、计算算法以及贝叶斯分析中的MCMC算法问题. 另外在实际生活中,特别在林学、经济学和生命科学等学科中异方差数据是广泛存在的,因此半参数双重回归模型有着广泛的应用背景,结合这些学科中的具体问题考虑半参数双重回归模型的实际应用.
中文关键词: 双重回归模型;变量选择;贝叶斯估计;半参数回归模型;纵向数据
英文摘要: With understanding of the real world in-depth, the actual data of our study is getting more complicated. If we just use simple statistical models to describe and study them, a lot of analysis have not been close to the actual results. Therefore, it is necessary for us to use the more complex and comprehensive models to describe these complex phenomena. Semiparametric double regression models is one of these classes. The main feature of this model is reflected in the attention of the variance. However, most literature focused on the mean regression models. Based on variance parameters (or dispersion parameters) given a model structure, statistical inference for the double regression models is rarely studied. However, double regression models have received a lot of attention in recent years. So, this project intends to study statistical inference problems for the semiparametric double regression models with complex data, for example, variable selection and Bayesian estimation. We will focus on the theoretical property and computational algorithms of variable selection for semiparametric double regression models with complex data, and MCMC computational algorithms in the Bayesian analysis. In addition, in real life, especially in forest science, economics and life sciences and so on, heteroscedastic data is also
英文关键词: double regression models;variable selection;bayesian estimate;semiparametric regression models;longitudinal data