The Research On The Diagnosis And Application Of The Variable Structure Point Based On Copula Function

With the development of world integration,China's foreign trade economy is increasingly open,and international financial crises are frequent,which has an impact on China's national economy.Therefore,measuring financial market risks is particularly important.Copula theory can measure the nonlinear relationship between random variables,and can consider the edge distribution function and joint distribution function of each sample hierarchically,which becomes an effective tool for studying financial time series.In this paper,the characteristics of Copula theory and the distribution characteristics of peak and thick tails of financial time series are synthesized.The most suitable binary normal Copula-GARCH(p,q)-t model is selected,and the nonlinear time-varying correlation coefficient of Copula function is used.Study the correlation changes of time series.Based on the traditional variable structure point test and Z test,the method of truncating interval variable structure point t test is proposed,and the deviation degree is measured by calculating the k-th order central moment to further locate the variable structure point,and the Granger causality test is used to test the industry.The direction of fluctuation and the size of the fluctuations have a certain practical significance for the study of financial time series.This paper selects the daily closing prices of the four sets of SZSE industry indices from July 30,2010 to March 22,2019,namely the mining index(399232),the hydropower index(399234),and the building index(399235).The Cultural Index(399248)was used as the study sample data for empirical analysis.After comprehensive consideration,we choose the binary normal Copula-GARCH(p,q)-t model,and fit the sample data to obtain the time-varying correlation coefficient and the constant correlation coefficient of the industry index return rate series.The central moment is measured to diagnose the existence of the variable structure point,and the direction of the change is verified by the Granger method.The empirical results show that: First,the time-varying correlation coefficient can be more sensitive and accurate than the constant correlation coefficient to measure the correlation between random variables.Secondly,the variable structure point diagnosis method based on the truncated sample t test and the Chow test are compared with the traditional Z test method.The scope of application is wider and the rate is faster.The more detailed the sample is divided,the more accurate the diagnosis result is,and it has certain reference value for the future research on financial market risk measurement.