Nonlinear Granger Causality Test And Its Application In Quantitative Analysis

Quantitative analysis is a general term,which mainly refers to methods and strategies that are related to quantitative trading and quantitative investment.It uses statistical analysis and mathematical models to replace human subjective judgments,selecting a variety of ‘high probability'events that can bring excess returns from a large amount of historical data in order to develop trading strategies and investing technologies.At present,the idea of quantitative analysis has gradually become a focus of domestic financial research.Among plenty of quantitative analysis problems,the causality in data is the focus of quantitative research.This paper analyzes the RMB exchange rate through a nonlinear Granger causality test,which has some certain value in guiding for exchange rate investors.Firstly,we review Ganger causality test and deepen it.Existing research manifests that linear Granger causality test cannot test the nonlinear causality in data.Nevertheless,the nonlinear Granger causality test has various types and is able to solve complex problems,so it is expected to be further deepened and expanded.For the nonlinear Granger causality test,the density estimation method using the explicit function in the nonlinear Diks-Panchenko causality test is used for reference.This paper uses the kernel function density estimation method to give an approximate test statistics of NKTN and discuss its limit properties.The feasibility of nonlinear test statistics of NKTN is studied by numerical simulation,and the correlation analysis,unit root test and nonlinear test are performed on the empirical data.Secondly,in order to further discuss the testing effects of nonlinear test statistics of NKTN and its limit properties,this paper conducts empirical research by comparing linear Granger causality test and nonlinear NKTN causality test.The results indicate that after examining the original data of RMB exchange rate and the residual data that filtered through the VAR model,GARCH-BEKK model and APARCH model,the linear Granger causality and nonlinear Granger causality among eight kinds data of RMB exchange rate can be obtained.At the same time,we analyze the similarities,differences and possible causes of the different test methods on the corresponding issues based on the test results.We find that the factors affecting the change of causality are more than fluctuation effects and spillover effects,which means that this issue needs to be further studied.Finally,on the basis of nonlinear Granger causality test,this paper uses the methods of reducing errors proposed by Diks and Wolski for reference,studying the impact of data sharpening method in the multivariate nonlinear Granger causality test.Meanwhile,this paper gives the corresponding multivariate nonlinear test statistics of QNKTN and discusses its limit properties.Through the numerical simulation and empirical analysis,the feasibilities of data sharpening method and nonlinear test statistics of QNKTN are verified.According to the empirical research,nonlinear NKTN test,multivariate nonlinear QNKTN test as well as their limit properties are summarized.The existing problems in the testing methods and techniques are also pointed out,which indicate the direction for the next theoretical research and improvement of the testing methods.