基于倾斜-t Copula的半参数马尔可夫链

在不指定时间序列结构的情况下,我们的分布模型是基于多变量离散时间的相应马尔可夫族和相关变量一维的边际分布.这样的模型可以同时处理时间序列之间的相互依赖和每个时间序列沿时间方向的依赖.具体的参数copula被指定为倾斜-t. 倾斜-t Copla能够处理不对称,偏斜和粗尾的数据分布.三个股票指数日均收益的实证研究表明,倾斜-t copula的马尔可夫模型要比以下模型更好:倾斜正态Copula马可夫, t-copula马可夫, 倾斜-t copula但无马尔可夫特性.

Skew-t Copula-Based Semiparametric Markov Chains

Without specifying the structure of a time series, we model the distribution of a multivariate Markov process in discrete time by the corresponding multivariate Markov family and the one-dimensional flows of marginal distributions. Such models tackle simultaneously temporal dependence and contemporaneous dependence between time series. A specific parametric form of stationary copula, namely skew-t copula, is assumed. Skew-t copulas are capable of modeling asymmetry, skewness, and heavy tails. An empirical study with unfiltered daily returns for three stock indices shows that the skew-t copula Markov model provides a better fit than the skew-Normal copula Markov or t-copula Markov model, and the skew-t copula model without Markov property.