Study On Autocorrelation Test Of Error Sequence Based On Monte Carlo Simulation

Due to the"inertia"formed by the continuity of economic development,the problem of"autocorrelation"is often encountered when using the time series data to establish the econometric model,that is,the values of each period of the random error term ?_t in the model are related to each other.The existence of autocorrelation will increase the estimation error of the model coefficients,and reduce the reliability of the statistical test and the accuracy of the prediction.Therefore,when carrying out econometric analysis,it is necessary to check whether the model has autocorrelation.However,several commonly used autocorrelation test methods have some problems in different degrees,which is of great significance for further research.The main research work of this paper is as follows:(1)For the given explanatory variables,Monte Carlo simulation method can be used to obtain the simulated critical values of DW test,Durbin's h test and regression test.Therefore,this can overcome the defects of uncertain regions in DW test,inaccurate critical values in Durbin's h test and no clear critical values in regression test.(2)According to the simulated critical value,the test power of the DW test,Durbin's h test,and regression test was estimated by Monte Carlo simulation.The results showed that the test power of the ? test and the t test was significantly greater than the DW test,and also superior to the Durbin's h test,except that the test power was very close to one.Therefore,the ? test and the t test can be used as an alternative to the DW test and the Durbin's h test.(3)For the test of high-order sequence correlation,this paper proposes LM test without explanatory variables(lm test)and F test.For the given explanatory variables,the simulated critical values of LM test,F test and lm test can be obtained by Monte Carlo simulation.It can overcome the defects of inaccurate critical values in LM and F test.It can also overcome the defect that lm test does not have a clear threshold.(4)According to the simulated critical value,the test efficacy of LM test,F test and lm test was simulated.The results showed that the test power of LM test and F test was basically the same in any case.When the sample size was not more than 20,the test power of the lm test was better than the other two.When the sample size was more than 20,the test power of the other two methods was superior to the lm test in most cases.(5)For a given explanatory variable,the simulation critical value of the-statistic of the highest order delayed regression coefficient in the LM test can be obtained by Monte Carlo simulation.Therefore,the inaccurate defect of the highest order lag regression coefficient t-test in the LM test can be overcome.