Statistical Inference Of Granger Causality With A Semiparametric Quantile Regression Analysis

Granger causality has become an important concept and tool in the study of the dynamic linkages between economic variables and to explore whether or not an economic variable helps forecast another one,it is also a focus and hotspot in the research of the theory,researchers have suggested a variety of methods to test the existence of Granger causality in the literature.In the thesis,the main objective is to develop a class of semiparametric time series quantile regression models that are of partially linear structures,the values of some coefficients may be functions of informative covariates,we propose two estimation procedures of both parametric and nonparametric functional coefficients,one is estimation procedure of semiparametric time series quantile regression models that are of partially linear structures with statistical theory established under a more general framework of near epoch dependent(NED)data generating processes,the other estimation procedure of a support vector quantile regression model with partially varying coefficients,three stage approach and the iteratively reweighted least squares procedure.The reasons why we will adopt such a semiparametric time series quantile regression models that are of partially linear structures are not only because it is a natural extension of the structure of linear Granger causality analysis but also because it enables us to estimate and hence understand the likely structure of the Granger causality if it exists.Finally,we have empirically applied the proposed methodology to examine the existence or effects of Granger causality between each pair of the financial markets involving China,USA and UK.From these results,we can see that the USA and UK stock prices 1 week ago help to predict current China stock market behaviors.