Unit Root Test Based On Bayesian Method

The unit root test is an important method to check whether the time series is stationary.The classical unit root test method is very popular among scholars.As the considerable efforts of the statistics in this field,they found that the classical unit root test method was inadequate and proposed that bayesian theory could solve the problem.Therefore,bayesian unit root test has become a research hotspot.In this paper,the unit root test of normal distribution time series and heavy-tailed time series is studied.In the normal distribution time series,the test result is influenced by the prior information of the autoregressive coefficient.In order to verify the existence of unit root in time series more effectively,this paper constructs a mixed prior distribution function to make up for the disadvantages of excessively rejecting the unit root hypothesis in bayes test,the validity of this method is proved by Monte Carlo simulation and example.In addition,There is less research on unit root test in heavytailed time series,we study the existence of unit roots in heavy-tailed time series by bayesian method.Through Monte Carlo simulation experiment,the influence of freedom and prior information on unit root test results is studied.Finally,the feasibility and usefulness of this method are proved by concrete example.