Exploration And Application Of Serial Correlation Test Method For Quantile Predictive Regression Model

In recent years,compared with the traditional regression prediction model,some scholars have proposed the quantile predictive regression(QPR)model,which can reveal more information,and its research results are more stable.However,when applying the model,it involves the order determination of time series,because only when the correct model order is determined,the subsequent statistical inference processes such as parameter estimation and hypothesis testing are meaningful.On the contrary,when the model order is misplaced,it will result in sequence autocorrelation in the residual sequence.In the financial market,stock return prediction has always been the research focus of researchers in this field and the topic of investors’ concern.How to predict the changes of stocks in the complex stock market and select high-quality stocks with high return on assets,these questions are not only of great significance to asset pricing and asset allocation,but also can carry out risk factor analysis and corresponding control and prevention of market risks.In view of the scarcity of research literature related to this model,this theory mainly discusses the residual series correlation of quantile predictive regression(QPR)model.Therefore,this thesis studied the serial correlation test of quantile predictive regression(QPR)model,and applied the research method to the actual data,which is conducive to enriching the prediction research of stock returns.It not only enriches the theory of serial correlation test,but also is conducive to the development of the research on the prediction of stock returns.In terms of research methods,this thesis constructed test statistic applicable to quantile predictive regression(QPR)model based on Portmanteau test and made Q-Q plot of its asymptotic distribution.Meanwhile,the robustness and superiority of the proposed method have been further verified by numerical simulation experiments.In practical application,this thesis applied this test to prediction of the weighted excess stock return in the CRSP database,for testing its serial correlation.The research shows that this new test method has a good test effect under the quantile prediction problem.Furthermore,in the test of sequence correlation in the stock market,there may be differences in sequence correlation at different quantiles,which is a phenomenon that may be overlooked in traditional mean regression.If the quantile predictive regression(QPR)model has correlation under a certain quantile,it suggests that the corresponding quantile needs to be re modeled.The research results demonstrate the significance and value of this testing method from the perspectives of simulation and practical application.The existing literature has not yet involved the serial correlation test under the quantile regression(QPR)model.Based on portmanteau test,this thesis deduces the sequence correlation test method of this model,which not only enrichis the theory of sequence correlation test,but also applies the research to the financial market.It is beneficial to the use and development of quantile prediction regression(QPR)model in financial markets.