非参数随机条件持续期模型及其迭代算法

发布时间：2018-04-23 03:15

本文选题：SCD模型 + 持续期　；参考：《统计研究》2011年08期

【摘要】：随机条件持续期(SCD)模型能有效刻画超高频时间序列中持续期的变化,但该模型假定期望持续期生成机制固定,且模型参数估计存在一定的困难。文章在不假定条件均值形式和冲击项分布的基础上结合核估计方法提出了非参数SCD模型及其迭代求解方法。然后,基于TEACD(1,1)模型生成的模拟数据,将非参数SCD模型与用卡尔漫滤波进行伪似然估计的参数SCD模型和用Gibbs抽样进行马尔科夫蒙特卡罗估计的参数SCD模型的拟合效果进行比较,实证表明在大样本条件下非参数SCD模型的拟合效果与用MCMC估计的参数SCD模型的拟合结果相差不大,但明显优于用QML估计的参数SCD模型的拟合结果,且非参数SCD模型能为参数SCD模型的参数设定提供参考。
[Abstract]:The stochastic conditional duration (SCD) model can effectively describe the duration variation in UHF time series, but it assumes that the expected duration generation mechanism is fixed, and the model parameter estimation is difficult. In this paper, the nonparametric SCD model and its iterative solution are proposed based on the kernel estimation method and the conditional mean form and shock term distribution are not assumed. Then, based on the simulated data generated by the tea model, the fitting results of the non-parametric SCD model are compared with that of the parametric SCD model using the Karman filter and the parametric SCD model using the Gibbs sampling to estimate the Markov Monte Carlo estimation. The empirical results show that the fitting effect of the non-parametric SCD model under the condition of large sample is not different from that of the parametric SCD model estimated by MCMC, but it is obviously superior to the fitting result of the parameter SCD model estimated by QML. And the nonparametric SCD model can provide reference for parameter setting of parametric SCD model.
【作者单位】：东南大学经济管理学院;
【基金】：国家自然科学基金项目“流动性调整期望损失La-ES和最优变现策略”(70671025)资助;国家自然科学基金项目“基于复杂网络的银行间传染性风险及其演化模型研究”(71071034)资助
【分类号】：F224;F830.9

【参考文献】