The Method And Application Of Quantile Unit Root Test

Unit root test is an important method to determine the stability of time series,and the stability is the premise of many time series modeling.Therefore,unit root test has become a work before modeling of time series.Traditional unit root test methods are mostly based on the assumption of many ideal conditions.For example,linear trend,the residuals obey Normal distribution,there is no auto-correlation of residuals,These assumptions are not satisfied,the traditional unit root test method can not be used.Therefore,considering the superiority of the quantile regression method,we can use the quantile unit root test method to analyze and modeling.Quantile unit root test method can test the different points of the statistics,and found the existence of the unit root in different quantile points.Quantile unit root test overcomes the traditional single test method,At the same time,the assumption that the quantile units are satisfied is much more relaxed than the traditional method.In this paper,by using the traditional unit root test method to analyze,at the same time,it also carries out the analysis of the quantile unit roottest method,By comparing the use of different methods to find the most suitable method.This is also in order to better illustrate the advantages of quantile unit root test method.Through Monte Carlo simulation,the simulation results show:When the random error term follows the normal distribution,The ADF unit root test method is consistent with the PP unit root test method and the quantile unit root test method,The first two methods are better than those of the third,When the random error term obeys the t distribution with degrees of freedom of 2 and 3,The method of quantile unit root test is better than ADF unit root test and PP unit root test,In the empirical analysis,this paper uses the data of the annual consumer price index of China from January 1990 to November2016,Results show that Under different unit root test methods,the time series data is non-stationary or stationary,By using the method of quantile unit root test,The time series data have different test statistic values at different points,Therefore,the test results also change with the change of the site,This reflects the distribution characteristics of forecast variables more comprehensively.So as to make a better decision.In order to be persuasive,In this paper,another empirical analysis based on Provincial data,The empiricalanalysis is based on the monthly data of the consumer price index of 31 provinces in China from January 1995 to November2016.The empirical results show that the ADF unit root test and the PP unit root test of the consumer price index of 31 provinces and cities in china,Only the unit root test in Chongqing can not reject the unit root test,and the other 30 provinces and cities are rejecting the original hypothesis.The results of the quantile unit root test showed that the results of unit root test were significantly different under different quantiles,At the same time,there is a rejection of the original hypothesis and can not reject the original assumptions.The test result is more comprehensive and accurate.