金融指数的单变点非参数极大似然估计研究

Study on Nonparametric Maximum Likelihood Estimation of Single Change-point for Financial Index

对于独立数据的变点估计方法的研究有很多,然而对于金融领域常见的自相关时间序列数据的应用却不明确。文章首先介绍了一种用于探测数据序列变点的非参数方法,继而提出了一种改进步骤使之适用于自相关的金融时间序列数据。统计模拟结果显示,所提出的改进方法能得到更加一致的估计变点。最后,通过对2015年4月7日至2017年4月6日的上证综合指数进行变点的判别,并得到稳健估计,由此说明对于自相关的金融时间序列数据,该改进方法具有必要性和适用性。

There are many researches on variable point estimation method of independent data, but the application of autocorrelation time series data in financial field is not clear. This paper first introduces a nonparametric method for detecting the variable points in data sequences, and then proposes an improved step to make it applicable for autocorrelation financial time series data. The statistical simulation results show that the proposed improved method can obtain a more consistent estimation of the variable points. Finally, the paper obtains a robust estimate by identifying the variable points of the Shanghai Composite Index from April 7, 2015 to April 6, 2017, which indicates that the improved method is necessary and applicable for autocorrelation financial time series data.