基于时序图模型结构估计的股票联动研究

我国股市个股价格同时上涨或同时下跌的联动现象极为普遍,传统上使用向量自回归、协整、有向非循环图等方法主要用于少量股票或市场之间的联动性研究,不适于直接对大规模个股之间的联动关系进行研究。文章关注大规模时序图模型结构建立及估计方法,通过将ADL方法引入SPACE算法,提出了可以估计高维低样时序图模型的ADL-SPACE算法;设计模拟实验考察了算法中惩罚参数λ值的设置对于节点自回归相关性捕获的有效性;在实证研究中,文章使用了ADL-SPACE算法对个股联动研究了三方面的内容:1.基于个股联动的代表性行业之间的联动性;2.设计了我国A股市场中行业联动强度,对行业内外联动性进行综合评价和分析;3.采用一阶滞后个股基于时序图模型结果构造了投资组合,模拟显示收益预期表现良好。以上研究均表明时序SPACE图模型方法在大规模股票的联动探测中有较好的应用前景。

Co-movement Analysis for Stocks Based on Time Series Graphical Models Structure Estimation

Co-movement which Stock prices rise together or fall together is popular in Chinese stock market.Classical methods,such as VAR,Co-integration,DAG,are limited when they are directly applied to co-movement analysis with huge stocks.In this paper,we focus on building graphical structure and estimating for analyzing co-movement with high dimensional small sample time series.By introducing ADL into SPACE model,we propose ADL-SPACE that can estimate high-dimensional-time-series Graphical model.We design simulation experiment to explore the efficiency of ADL-SPACE algorithm in different penalty A setting.In the demonstration research,this paper apply ADL-SPACE algorithm to construct strengthen index to rank the strength of Co-movement with industry plates in China’s stock market.We construct a good strength index Portfolio based on the result from the ADL-SPACE model which prove ADL-SPACE algorithm is useful in practice.