基于混频模型的高阶矩最优因子个数识别研究

研究目标:在构建包含偏度和峰度的高阶矩投资组合情况下,为了减少设定误差的同时进一步降低传统朴素估计存在的较高估计误差,本文构建了混频多因子模型并提出了一种新的包含高阶矩条件下最优因子个数识别方法,并探讨了其适用性。研究方法:在理论上分析了基于混频多因子模型得到统计量的渐进性质,并通过蒙特卡洛模拟在有限样本条件下进行了检验。研究发现:新方法可以更为准确地识别包含高阶矩条件下最优因子的个数,相对于基于信息准则构建的高阶矩因子个数识别方法具有明显的优势。研究创新:基于混频模型方法,进一步提高因子对模型的解释能力,在混频模型设定适当的条件下,由扰动项构建得到的高阶矩矩阵应具有明显的稀疏性特征,利用这一特征识别最优因子个数,从而进一步降低高阶矩矩阵的估计误差。研究价值:通过使用混频模型,在提高对模型解释能力的同时,构建一种相对于传统信息准则方法更能准确识别包含高阶矩信息的最优因子个数估计方法。

Determining the Number of Factors of Higher-Order Co-moments Based on MIDAS Model

Research Objectives:In order to reduce the high estimation error of traditional naive estimation in the case of constructing high-order co-moments portfolio with skewness and kurtosis,this paper proposes a new method of determining the optimal number of factors when high-order co-moments are considered based on MIDAS model.Research Methods:In theory,the asymptotic properties of the statistic obtained by the new method are analyzed and tested by Monte Carlo simulation under finite sample conditions.Research Findings:The new method can more accurately identify the optimal number of factors including high-order moments,and has obvious advantages over the method based on information criterion method.Research Innovations:Based on the MIDAS method,the interpretation ability of factor model is further improved,and under the appropriate conditions by the MIDAS model,the high-order co-moments obtained by the error term should be sparse,which could be used to identify the optimal number of factors,thus further reducing the estimation error of the higher-order co-moments matrices.Research Value:By using the MIDAS model,while improving the interpretation ability of the model,a method of estimating the optimal number of factors containing high-order moment information is constructed and is proven more accurate than the traditional information criterion method.