新息修正偏态分布的LN-ACD与LN-LOG-ACD模型及参数估计

发布时间：2018-04-25 06:28

本文选题：ACD模型 + 新息　；参考：《系统工程》2011年10期

【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。
[Abstract]:In order to satisfy the necessary conditions of the innovation distribution of ACD model and the requirement of solving LOG-ACD model with non-pseudo-likelihood Kalman filter, a modified skewness distribution is proposed on the basis of skewness distribution. Based on this distribution, a LN-ACD model is constructed. According to the moment condition, the parameter constraints are obtained, and it is proved that the LN-ACD model has ARM A property and its parameters can be estimated by the maximum likelihood method. In order to avoid the constraint on the parameters in the LN-ACD model, the LN-LOG-ACD model is further proposed, and it is also proved that the logarithmic durations characterized by the model have ARM A property. Moreover, the disturbance term of the state space equation transformed from the LN-LOG-ACD model is followed by Gao Si process. The Karl-diffuse filtering solution of the parameters is a non-pseudo-likelihood estimate, and the solution is effective. Finally, based on three groups of simulated data from different innovation distribution, this paper demonstrates the excellent characteristics of LN-LOG-ACD model from different angles.
【作者单位】：东南大学经济管理学院;
【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
【分类号】：O212.1;F830

【二级参考文献】