Outlier Change Point Estimation Of Mean And Variance Of Panel Data

Panel data includes two parts:time dimension data T and cross-sectional data N.It is very important in econometric and financial science research and learning,in which change point analysis is an important part of panel data research.The change point problem is closely related to people's daily life and has attracted the attention of scholars from all walks of life.Change point detection and estimation are widely used not only in the field of quality control,but also in many scientific fields,such as epidemiology,meteorology,finance,geology,medicine,image recognition and so on.Two basic and important aspects of panel data change point analysis are mean change point analysis and variance change point analysis.In econometric analysis,for panel data Yit,the mean value is usually used to measure the average return of assets,and the variance is used to reflect the risk fluctuation of market economy.Based on the research of Bai(2010),this paper studies that when T and N are large enough,the mean change point ?₀ in the panel data model and variance change point ?₀ change point estimation when is ectopic.Firstly,it is assumed that there is a single same change point of mean and variance(?₀>?₀)between the N sequences in the panel change point model Yit;Then,LSE is used to estimate the mean change point when the variance change point appears before mean change point in the panel data,and M-C numerical simulation is carried out;Finally,based on the mean change point estimation,a parameter adjustment CUSUM cumulative sum estimator is proposed to estimate the panel variance change point,and M-C numerical simulation is carried out.Through the above panel change point model change point estimation method,we give the change point estimators of panel mean single change point and variance single change point respectively,and prove the relevant convergence speed and consistency of the estimators M-C numerical simulation shows that when the variance change point appears before the mean change point,using LSE to estimate the mean change point is also applicable;at the same time,when the parameter CUSUM cumulative sum estimator proposed in this paper estimates the panel variance change point,the estimation performance is better when the parameter ? in[0,0.7).Comprehensive analysis of the estimation methods applied and proposed in this paper can approach the real change point with probability 1,which can provide some theoretical support for the research of panel data change point in financial market.