Non-stationarity Data Test And Estimation Of Spatial Error Correction Model With Bayesian

With the development of econometrics,especially the unit root process of time series and the development of cointegration theory,it brings a revolution to the development of mainstream econometrics.Time series analysis has established a relatively perfect theoretical analysis framework and achieved a good application.As a branch of econometrics,spatial econometrics has the unique advantage in dealing with the spatial interaction(spatial autocorrelation)and spatial structure(spatial heterogeneity)of cross-sectional data and panel data,with the help of a classical model for example spatial error model,spatial lag model and geographic weighted regression model.The combination of spatial econometrics and geographic information system(GIS)is widely used in the fields of real estate economy,environment and resource economy,but the development of spatial non-stationarity econometrics is not perfect in spatial econometrics.Based on the unit root and cointegration theory of time series,this paper briefly introduces the research methods of time series unit roots,and then introduces the development of spatial non-stationary econometrics in recent years,including the spatial unit root and spatial cointegration theory of the first proposed and its test methods,spatial unit root process data generation methods and testing,as well as spatial error correction model of different forms and parameter estimates.Similar to the time series of DF tests,the non-stationary test of the critical value is also proposed.Moran's I is a method to the agglomeration effect of variable,and this paper also introduces a study of the relevance of two spatial processes.The study of unit root of spatial panel data has mature theory,but the study of the stability of spatial cross section data is not perfect,Therefore,the research of this paper is based on spatial cross section data.The innovation of this paper is that test the rationality of the critical value of non-stationary test by time series method and suggest to test the smoothness of spatial data by two-step Lagrangian test.Two classes of non-stationary spatial processes stationary processes with trend items and random walks with drift items are examined.The simulation experiments can distinguish between the two different processes that cause non-stationary.In view of the fact that the error is large at the parameter boundary with the instrumental variable method estimating the spatial error correction model,Bayesian method is used to estimate the spatial error correction model.The results show that the Bayesian method is superior to the instrumental variable method in the whole,even if the the parameter boundary.