摘要
本文将高斯混合自回归模型(GMAR)引入到风险价值(VaR)的计算上,并将其用于计算上证综指收益的VaR,将所得结果与GARCH类模型进行比较分析发现,GMAR能捕捉剧烈波动的条件异方差性,但对于较小的波动无法捕捉,因此所得VaR曲线较GARCH类模型平坦。而在不同的显著性水平下,GMAR模型预测VaR的能力都显著优于GARCH类模型。
The paper introduced Gaussian Mixture Autoregressive Model (GMAR) to the calculation of Value-at-Risk (VaR), calculated the VaR of the return of Shanghai Stock Exchange (SSE) composite index. By comparing the results with that of Generalized Auto-Regressive Conditional Hetero-skedasticity (GARCH), it is found that GMAR can capture heteroscedasticity's sharp fluctuations, but not the minor ones, so that the VaR curve of GMAR is smoother than GARCH's. Besides, under different significance levels, GMAR's capability of predicting VaR is obviously better than that of GARCH.
出处
《浙江社会科学》
CSSCI
北大核心
2013年第5期48-55,75+156,共10页
Zhejiang Social Sciences
基金
国家自然科学基金(基金号71101126)"基于自组织聚类半参数系统的金融时间序列预测模型研究"的阶段性成果
浙江工商大学研究生科研创新基金资助的成果