Estimation And Application Of Generalized Spatial Variable Coefficient Ridge Regression Model

With the increasingly development of science and technology,on the specific ge-ographic location data can be collected by the discrete space more show the charac-teristics of high dimension and high complexity,and the data variables or inner link,more or less so between spatial data variable linear inevitable.Common processing method of generalized spatial data are Geographically Weighted Regression(GWR)and Local Linear Geographically Weighted Regression(LLGWR),and the above two methods of estimation in data when there is a certain degree of collinearity between variables,which can lead to ultimate estimate result unreliable;in addition,local geographical weighted linear estimation process will also introduce the collinearity between variables,the local geographical weighted linear face stronger collinearity.According to the above problem,this paper based on variable coefficient model of generalized space Geographically Weighted Ridge Regression(GWRR)method and the Local Linear Geographically Weighted Ridge Regression(LLGWRR)method.First of all,based on the generalized space variable coefficient regression mod-el,using geographically weighted ridge regression estimation method of generalized spatial variable coefficient to solve the parameters in the regression model.Then,ac-cording to the simulation experiment coefficient function under different collinearity data simulation results,according to estimate stability coefficient function metrics,and geography,comparing the weighted coefficient estimation results to verify the present strong collinearity between variables,geographical weighted coefficient of ridge regression estimation is geographically weighted estimate coefficient more ef-fectiveness;again based on geographical weighted ridge regression local geographical weighted linear ridge regression estimation method,data simulation and compar-ing with the results geographical weighted estimation of ridge regression method,to explore the local geographical weighted linear ridge regression is geographically weighted ridge regression in the process of linear space data under the estimation accuracy and estimation of the stability of the difference,find local geographical weighted linear ridge regression method has higher estimation accuracy and stabili-ty.Finally,based on the data simulation results it is concluded that the geographical weighted ridge regression estimation methods and local linear geographical weighted ridge regression estimation methods for treatment of linear space data with valid-ity and stability,in the comparison between the method of geographical weighted ridge regression estimation and the method of local linear geographical weighted ridge regression estimation,the method of local linear geographical weighted ridge regression estimation has some advantages.Respectively in this paper using geographically weighted ridge regression and partial linear geographical weighted ridge regression analysis in 2016 China's 31 provinces,municipalities and autonomous regions of tuberculosis incidence space changes,to explore the data presentation under the multi-collinearity passenger traffic,consumption level,regional GDP per capita,population density and per one thousand medical institutions of beds the five macro factors that affect the incidence of tuberculosis,the spatial variation characteristics of these effects were analyzed,and explore the ridge regression in dealing with multi-collinearity play a role in spatial data,finally comparing the result of the two methods of estimation.