A Structural Mutation Study Of Variance In Panel Data

Panel data is different from the time series data and cross-sectional data in one-dimensional data.It is two-dimensional data.If the time series data of an individual at different moments is regarded as a record,then panel data means that multiple records are selected at the same time,that is,multiple individual sample time series data are selected at the same time.From the cross-sectional direction,the panel data is expressed as the observation value of multiple individual samples at the same time,while from the longitudinal cross-sectional direction,the panel data is a time series.Panel data is often used as a research object in the field of economics,such as the population change of various provinces and cities in the country during a certain period of time.In the entire social development process,for example,changes in monetary policy will cause fluctuations in stock market prices,causing a set of data to have significant differences before and after a certain point.This is the break point.In the 1960 s,the problem of change point has been discovered and explored.Whether in mathematics or statistics,economics,finance,medicine,meteorology,etc,the problem of change point in data structure is a common research topic,and there are a lot of research topics.Practical application.There are three main directions for the study of change points.One is the problem of change point detection,the other is the problem of estimating the number of change points,and the other is the problem of estimating the location of the change point.The research of single change point and multiple change point in single time series,single change point in panel data are mostly,and most of them are to explore the mean change point.Nowadays,there are more and more researches on variance change point,the change in variance mainly reflects the fluctuation of data.This article mainly studies the position parameter estimation of the variance change point of panel data,that is,there are multiple variance change points in panel data at the same time.Sometimes there are not only change points in the variance,but also the mean value.Therefore,this paper also considers the problem whether the mean also has multiple change points and the location of variance change points is the same or not.When only the variance has multiple change points and the mean and variance have common multiple change points,the quasi maximum likelihood(QML)estimation method is used to estimate the location of the change points.When the mean and vari-ance have different multiple change points,the least square estimation method is used to estimate the mean change points first,and then the data is centralized.Then,the quasi maximum likelihood estimation method is used to estimate the multiple change points of the variance.The results show that the estimates obtained under certain conditions are consistent.In order to verify the accuracy of the conclusion,Monte Carlo simulation is used in this paper.The simulation results are consistent with the conclusion,which shows that this estimation method is effective.Finally,this paper makes an empirical analysis based on the monthly closing price data of 49 enterprises from January 2007 to June 2017,finds out the position of the change point and analyzes the reasons.