Comparative Research Of GARCH Models Based On Different Distributions

In recent years,there have been more and more online transactions in the financial market.This online transaction has both advantages and disadvantages.On the one hand,it makes trading more convenient and fast,on the other hand,it makes stock market volatility more and more intense,thereby increasing the risk of the stock market.Therefore,people are more concerned about how the volatility of the stock market changes,and on the basis of grasping the characteristics of financial asset price volatility,reduce risks and maximize returns.In practical applications,modeling and analyzing financial data through the volatility model can help investors better understand the laws of asset price fluctuations in the financial market,thereby minimizing losses,and effectively measuring and preventing financial market risks.How to choose a suitable volatility model and distribution form is a key step to improve the estimation accuracy of financial asset prices and effectively avoid risks.In terms of volatility model selection,this article takes the high-frequency volatility model-Realized GARCH model as the research object,which combines the realized volatility with the GARCH model,so as to make full use of the information of high-frequency data and overcome low-frequency fluctuation defects of the rate model.In terms of distribution form selection,this paper selects four commonly used “peaky thick tail skewness” distributions: T distribution,ST distribution,GED distribution,and NIG distribution,taking normal distribution into account.First,the basic GARCH model and its extended model are introduced in detail,and then the models mentioned in the article are compared and analyzed.Secondly,the construction process and statistical characteristics of the realized volatility are introduced.Based on the realized volatility and GARCH model,the paper puts forward the development and development of the Realized GARCH model,and introduces the distribution choice of error terms in detail.Finally,the relevant theories of model comparison criteria are introduced.Based on the theoretical analysis,this article models and analyzes the volatility of the Shanghai and Shenzhen 300 Index.Select 5-minute high-frequency data for a total of 730 trading days from January 5,2016 to December 28,2018 from the CSMAR database,and set the yield error items to follow the normal distribution,T distribution,ST distribution,and GED distribution And NIG distribution.Through R language and Eviews correlation analysis,it is found that the CSI 300 index return series has the characteristics of sharp peak,thick tail and skewed distribution.Then estimate the parameters of the Realized GARCH model under five different distributions,and compare the Realized GARCHmodels under different distributions from the four aspects of model fitting ability,model distribution setting test,model volatility prediction effect and risk measurement analysis.The empirical analysis results show that the Realized GARCH model assuming that the error terms of the return rate follow the skewed thick-tailed skew distribution(T distribution,ST distribution,GED distribution,and NIG distribution)are better than the normal distribution,and can better fit the return Volatility,while the Realized GARCH model based on the NIG distribution performs best,and the Realized GARCH model based on the ST distribution performs second,so in terms of distribution selection,we can first consider the NIG distribution and secondly the ST distribution.Through research and analysis of the volatility of the Shanghai and Shenzhen 300 Index,investors in the stock market can have a deeper understanding of the short-term volatility characteristics of the Chinese stock market and have practical investment guidance value.