摘要
广义自回归得分模型以条件密度函数得分作为主要驱动,为我们提供了计算一个时变参数的框架。把传统的D藤结构拓展到时变D藤,并将其与广义自回归得分模型相结合来估计时变参数。对美元、欧元、日元、港元和英镑兑人民币的外汇汇率序列拟合边际分布,提取标准化残差序列并建立时变D藤模型,即对pair-copula分解式选择最优的时变copula族,最后通过仿真技术得到标准误差的箱线图。研究表明,该模型有效地刻画了变量间的相依关系,为描述金融资产收益率的波动率过程提供了一种新思路。
Generalized autoregressive score model use the score of the conditional density flmction as the main driver and provide a framework for estimating time-varying parameters. The traditional D-vine is extended to time-varying D-vine and used to evaluate time-varying parameters with GAS model. We. fit the marginal distribution about the foreign exchange rate series of US dollar, Euro, Japanese yen, Hong Kong dollar and Pound against the Chinese Yuan and then extract the standardized residual series and modeling the time-varying D-vine, that is, choosing the optimal family of time-varying copula for pair-copula function. Finally, getting the boxplot of standard errors by simulation technology. Research shows that the dependence among variables are usefully described by the model and this work provides a new method for describing the volatility of return of financial assets.
出处
《数理统计与管理》
CSSCI
北大核心
2018年第1期74-82,共9页
Journal of Applied Statistics and Management
基金
国家统计局重点课题(2014LZ41)
关键词
GAS
得分函数
时变D藤
外汇汇率
GAS, score function, time-varying D-vine, foreign exchange