| 吴吉林,孟纹羽.时变混合Copula模型的非参数估计及应用研究[J].数量经济技术经济研究,2013,30(8):124-136 |
| 时变混合Copula模型的非参数估计及应用研究 |
| Nonparametric Estimation in Time-varying Mixture Copula and its Application in Stock Markets |
| |
| DOI： |
| 中文关键词: 混合Copula 非参数方法 最优带宽 尾部相依性 |
| 英文关键词: Mixture Copula Nonparametric Method Optimal Bandwidth Tail Dependence |
| 基金项目: |
|
| 中文摘要: |
| 针对传统动态混合Copula参数建模所存在的缺陷，本文提出了混合Copula的非参数建模方法，即不对模型的参数进行任何形式的设定，而假设参数为时间的函数，运用局部极大似然估计法来对参数进行估计。我们推导了非参数估计量的渐近正态性质、讨论了估计量的偏差与方差的大小，并运用交叉验证法给出最优带宽的选择。有限样本下的蒙特卡洛模拟显示，时变混合Copula的非参数估计具有良好的统计性质。最后，将该方法应用于股市间尾部相依特征的研究，发现我国股市与美、日、港、英股市间的左右尾部相依性呈现明显的时变性和非对称性。 |
| 英文摘要: |
| On account of drawbacks in the parametric framework of modeling dynamic mixture Copula, this paper proposes nonparametric methods ,in which we don’t specify any concrete parametric forms for the parameters in mixture Copula, but just the function of time. We employ local maximum likelihood to estimate the parameters in Copula, then establish their asymptotic theories, the principles of choosing optimal bandwidth, and compare Monte Carlo simulations with different sample sizes. The result shows our model perform very well in finite sample size. Lastly,we apply the proposed models to measure tail dependecne in stock marktes, and shows there exist time-varying as well as asymmetric characteristcs in the left and right dependence between Chinese stock market and other main international stock markets. |
| 查看全文 |
