摘要
为将基于窗谱估计的模型验证技术应用于金融时间序列领域,以解决金融时间序列模型的设定正确性。本文通过实证比较的方法进行检验研究,结果表明:即使模型在5%置信水平下全部通过Basel规则、Kupiec统计量、Christofferson统计量检验,仍有可能存在模型设定错误,无法通过频谱分析模型验证法检验。也就是说,频谱分析模型验证法相对于当前商业银行或银监局在管理模型风险中广泛使用的回顾测试方法,在金融时间序列模型的正确性设定检验方面具有更强的模型验证能力,它能揭示系统的更全面的信息。
The window spectrum estimation technique was used to validate the financial time series model for bank finances. The empirical results show that even though the model is verified by the Basel accord, Kupiec statistics, and Christofferson statistics at the 5 ~ confidence level, the model may be wrongly specified and cannot pass the model specification test based on the window spectrum estimation. Thus, the window spectrum estimation technique provides more effective model validation than the traditional back-test methods usually used by commercial banks and monitoring agencies since it shows more information about the system.
出处
《清华大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2008年第9期1529-1532,共4页
Journal of Tsinghua University(Science and Technology)
关键词
金融
回顾测试
谱估计
模型验证
finance
back-test
spectrum estimation
model validation
