The Research And Application Of Nonparametric Quantile Regression Model For Panel Data

The parameter regression model is simple and convenient and widely used,but the simple parameter mean regression model cannot satisfy the complex data in life.On the one hand,because for the model,the parameter model is to make the assumption of the model in advance,given a specific model form,under certain assumptions,and then use the fixed model to analyze the data,while the actual data is complex and changeable,for the panel data,contains more information,the general parameter model can not meet the data requirements well.On the other hand,from the random error term distribution of the model,the existing model basically assumes that it obeys the norm Normal distribution,but for the error terms from the bias or peak state of the distribution there are defects.So based on the existing research,this paper further improves the panel data model based on some defects of the existing model.In order to explain the model in detail,the details are as follows:The first part discusses the nonparametric quantile regression modeling method for panel data in the framework of bayesian analysis.using low-rank thin-plate penalization spline expansion,the introduction of virtual variables and asymmetric Laplace distribution establishes a bayesian hierarchical quantile regression model,and gives Metropolis-Hastings sampling algorithms for unknown parameter estimation.The simulation results show that the new model has obvious improvement in unbiasedness and stability compared with the traditional model.In order to apply the model to the actual situation,the new model is applied to the rural residents of China In the demonstration of actual data of consumer expenditure and operating income,we analyze the relationship between net operating income and consumption expenditure under each consumer group,and draw a conclusion that the effect of consumption expenditure is positive stimulus with the increase of net operating income under each quartile,and this effect is more obvious at the high score.The second part,based on the research of the first part,continues to discuss the additive model quantile regression modeling method of panel data under the framework of Bayesian analysis.First,the nonparametric model is transformed into parametric model by the low rank thin plate penalty spline expansion and the introduction of individual effect virtual variables,and then the Bayesian hierarchical quantile regression model is established based on the assumption that the random error term obeys the asymmetric Laplace distribution.Through the decomposition of asymmetric Laplace distribution,the conditional posterior distribution of all parameters to be estimated is given,and the Gibbs sampling estimation algorithm for parameters to be estimated is constructed.Computer Simulation and Simulation The results show that the proposed method is obviously dominant in estimation robustness compared with the traditional mean regression method.Finally,we apply the quantile regression model to the study of the influence of aging,urbanization,economic factors and construction cost on house prices.It is concluded that the current aging has a certain suppression effect on house prices,while urbanization and the increase of income and cost make house prices show a large rise.