新息修正偏态分布的LN-ACD与LN-LOG-ACD模型及参数估计  

LN-ACD and LN-LOG-ACD Model and Its Estimation Based on the Improved Skewed Distribution of Innovation

作　　者：周伟[1] 何建敏[1] 余德建[1] 

机构地区：[1]东南大学经济管理学院,江苏南京211189

出　　处：《系统工程》2011年第10期97-102,共6页Systems Engineering

基　　金：国家重点基础研究发展计划(973计划)项目(2010CB328104-02);国家自然科学基金资助项目(71071034);江苏省普通高校研究生科研创新计划项目(CXZZ-01830184);教育部博士研究生学术新人奖项目

摘　　要：为了满足ACD模型新息分布的必要条件和LOG-ACD模型非伪似然卡尔曼滤波求解的要求,本文在偏态分布的基础上提出了一种修正偏态分布,基于该分布构建了LN-ACD模型,通过其矩条件求得参数约束,证明了LN-ACD模型刻画久期具备ARM A性,其参数可用极大似然法估计。为避免LN-ACD模型中对参数的约束,本文进一步提出了LN-LOG-ACD模型,同样证明了该模型刻画的对数久期具备ARM A性。并且,LN-LOG-ACD模型转化成的状态空间方程扰动项服从高斯过程,其参数的卡尔漫滤波解为非伪似然估计,解具备有效性。最后,本文基于三组服从不同新息分布的模拟数据,从不同角度实证了LN-LOG-ACD模型的优良特性。To meet the necessary conditions of innovation in the ACD model and non-pseudo-likelihood requirement by using the Kalman filter,this paper improves traditional skewed distribution.Then,LN-ACD model is constructed,whose no-conditional expected value,variance and the constraints of parameter are obtained.Of course,this paper proves the basic ARAM property in LN-ACD model just like the classic ACD model.To avoid the parameter constraints in the LN-ACD model,this paper further proposes the LN-LOG-ACD model,proves its ARAM property.For fluctuations in the state equation of the LN-LOG-ACD model is a Gaussian process disturbance,the solution is a non pseudo likelihood by using the Kalman filter.At last,this paper validates some good performance in this new model from different angels based on three simulation data produced by the ACD model using three kinds of distribution.

关 键 词：ACD模型 新息 修正偏态分布 卡尔曼滤波 参数估计 

分 类 号：F222[经济管理—国民经济]