A Research And Application Of The Quantile Regression Estimation Based On Bayesian

With the continuous research on quantile regression,the ideas of quantile regression and Bayesian quantile regression have proposed in the parameter estimation.Frequency analysis have generally used in the field of quantile regression parameter estimation.With the rise and development of Bayesian statistical school,combined with Bayesian estimation method and quantile regression model have formed a new research direction and widely used in economic,financial and other social fields.This paper focuses on the Bayesian estimation of quantile regression,and proposes two Bayesian quantile linear regression models,which were based on AL distribution and AEP distribution on the basis of quantile regression model.Due to the complexity of the distribution function,Gibbs algorithm or M-H algorithm have proposed for AL distribution and AEP distribution respectively,and the quantile regression model was explored based on Bayesian estimation principle.In order to test the effectiveness of the proposed method,Bayesian quantile regression models with different methods and distributions were simulated.The experiments showed that when the data show sharp peaks and thick tails,the Bayesian quantile regression estimation accuracy based on AL distribution was Effective.When the data was normal or abnormal,Bayesian quantile regression based on AEP distribution had strong robustness.Finally,in order to further test the effectiveness of the proposed method,we analyzed the influencing factors of grain production in Guizhou Province by using this method.The empirical results showed that the impact of the affected area of crops,the population involved in the employment of the primary industry,agricultural production expenditure,per capita grain production and the price index of agricultural means of production on grain production in Guizhou Province has significantly different at different quantiles.