摘要
本文讨论结构经济时间序列用状态空间模型进行分解处理的方法．在§１中综述结构时间序列的状态空间描述．§２中着重论述了将处理不完全数据的ＥＭ－算法应用于状态空间模型参数的极大似然估计．在§３中给出采用本文所述方法对一些我国宏观经济序列的计算实例．
A economic time series model can be decomposed directly in terms of the components of trend, seasonal, cycle and irregular, Each component is cha- racterized by unknown variance-white noise perturbed difference equation of AR model constraints. The constraints are expressed in state space model form. Kalman filter is used to compute maximum likelihood and estimate parameter. In our paper, indirect maximization of the likelihoop function via the EM algorithm is substituted for direct maximization of likelihood function via Newton algorithm, etc. Exmaples of economic time series in our country are given.
出处
《数理统计与管理》
CSSCI
北大核心
1992年第5期55-60,40,共7页
Journal of Applied Statistics and Management
关键词
状态空间
结构经济序列
EM算法
State space Structural time series seasonal adjustment likelihood Kalman filter EM algorithm Stochastic trend
