高频数据驱动的连续时间模型估计  

A High-frequency Data-driven Estimation Approach of Continuous Time Series Models

作　　者：赵驰[1] 殷炼乾 

机构地区：[1]西安交通大学经济与金融学院,陕西西安710061 [2]金禾经济研究中心,陕西西安710049

出　　处：《统计与信息论坛》2010年第5期18-21,共4页Journal of Statistics and Information

基　　金：国家985工程二期项目《中国宏观经济与金融预测平台研发》(07200701)

摘　　要：现代金融经济学中连续时间模型能够更方便地描述重要经济变量的动态过程如股价、汇率和利率等。为连续时间模型提出了一种高频数据驱动的二阶段估计方法,增强了连续时间扩展模型的弹性和可操作性。以Vasicek模型为例给出了该方法的应用实例,首先在第一阶段使用实现波动率方法估计出模型的扩散项参数,然后使用实际数据的稳态分布的前向方程估计漂移项参数。此方法对模型初始设定和优化算法依赖程度低,结果较为稳定可靠。In modern financial economics continuous time-series diffusion models are more convenient to deal with than discrete time models when they are used to depict important economic variables such as stock prices,exchange rates and interest rates.This paper proposes a two-stage method for continuous time-series diffusion models based on high frequency data.Compared to other computation-driven methods,our method is data-driven and very flexible in model specification. A practical example of Vasicek model estimation shows that first of all we use realized volatility to estimate the diffusion parameter of the model then forward kolmogorov equation of stable distributions is applied to estimate the drift parameter.This method relies less on initial setting of models and optimization algorithm,needs no initial values of input and delivers more reliable results.

关 键 词：连续时间模型 模型估计 高频数据 实现波动率 

分 类 号：F830[经济管理—金融学]