标题:Co-trading networks for modeling dynamic interdependency structures and estimating high-dimensional covariances in US equity markets 作者:Yutong Lu,Gesine Reinert,Mihai Cucuringu
对于一个线性因子模型:协方差可以用以下等式表示:参考Ait-Sahalia和Xiu (2017),上式右边两项可以由特征值及特征向量进行估计:其中第二项表示股票的特质收益矩阵,参考Ait-Sahalia和Xiu (2017),为了提高协方差估计的稳健性,可以对特质收益矩阵进行过滤,仅保留在某一分类方法下属于同一聚类的元素值。作者分别使用了基于GICS的固定分类及基于co-trading network的时变聚类法。使用基于前一交易日5分钟数据计算的协方差矩阵作为对于下一交易日的协方差估计,并测试全局最小方差组合的收益。以下是两个分类方法的对比,可以明显看出基于co-trading聚类作为分类的方法明显优于GICS,策略的表现更加稳健,夏普比率更高:参考文献Lu, Yutong and Reinert, Gesine and Cucuringu, Mihai, Co-trading networks for modeling dynamic interdependency structures and estimating high-dimensional covariances in US equity markets (February 18, 2023).Yacine Ait-Sahalia and Dacheng Xiu. "Using principal component analysis to estimate a high dimensional factor model with high-frequency data". In: Journal of Econometrics201.2 (2017), pp. 384–399.David Dekker, David Krackhardt, and Tom AB Snijders. "Sensitivity of MRQAP tests to collinearity and autocorrelation conditions". In: Psychometrika 72.4 (2007), pp. 563–581.