项目名称: 交叉分类累加方法与合并方法的多层统计模型理论及其应用研究
项目编号: No.71261004
项目类型: 地区科学基金项目
立项/批准年度: 2013
项目学科: 管理科学
项目作者: 刘殿国
作者单位: 海南大学
项目金额: 35.6万元
中文摘要: 为了解决多层模型和累加方法的多层模型难以分析交叉分类和离散型因变量的研究对象所有一层样本都过少的问题,建立交叉分类累加方法与合并方法的多层统计模型理论,并将其应用到中国香蕉产业组织发展的研究中。首先,利用数据累加与合并方法转化多层数据的表示形式。其次,分别建立交叉分类累加方法与合并方法的多层统计模型理论:研究系列模型(多层Logistic回归模型、多层泊松回归模型等)的表示形式;在限制性最大似然法(REML)、收缩估计法、惩罚性似然估计(PQL)等方法中通过模拟研究选择合适的参数估计方法;研究估计的结果怎样预测、评价原始数据。最后,应用系列交叉分类累加方法与合并方法的多层统计模型研究中国香蕉产业组织的发展。交叉分类累加方法与合并方法的多层统计模型理论将是多层统计模型理论、累加方法的多层模型理论以及灰色系统模型理论的新发展。对香蕉产业组织发展的研究,将得到系统化的香蕉产业组织体系的数量评价。
中文关键词: 多层统计模型;香蕉产业组织;社会嵌入性;能源效率;环境效率
英文摘要: It is difficult for multilevel statistical model and multilevel statistical model for accumulated method to analysis too few samples in all object of cross classifications and discrete response for 1-level.In order to solve the problem, multilevel statistical model for cross classified accumulated method and merged method will be established by we, and apply it to organizational development of Chinese banana industry. First, using method of data accumulated and merged for converting multilevel data representation. Secondly, multilevel statistical model theory for cross classified accumulated method and merged method will be established by we: the representation of series model (multilevel logistic regression model, the multilevel Poisson regression model etc)will be established by we, to choose proper parameters estimation method through simulation studies in the restricted maximum likelihood (REML) and shrinkage estimation and penalty quasi-likelihood (PQL)methods etc, how to predict and evaluate the original data by the estimation results. Finally, apply series multilevel statistical model for cross classified accumulated method and merged method to organizational development of Chinese banana industry. Theory of multilevel statistical model for cross classified accumulated method and merged method will be new
英文关键词: multilevel statistical model;banana industry organization;social embeddedness;energy efficiency;environmental efficiency