基于变窗宽下非参数回归函数加权核估计的大样本性质-概率论与数理统计专业论文.docxVIP

燕山大学理学硕士学位论文Abs仃act 燕山大学理学硕士学位论文 Abs仃act Nonparametric model is widely used in the practical problems，the reason is that the form of the regression function in the nonparametric model is free， and the limits to the random varime(X，y)are fewer．In the past several decades， this model is studied carefully by the researchers of statistics，and many productions are arrived in both academic fields and in applied fields． Inthispaper，we supposethat(X，Y)isRxR r．v．andEl，1≤00，(Xl，y1)， (X2，Y2)， ，(z Y)are idd．samples from the population(X，y)，m(x)is a Bllknown unvariate function that satisfy： Y，=m(X，)+s，，i=1，2， ，H E(8，)=0， 仃2(X，)=Var(E，)oo There search object of this paper is to estimate m(x)using the samples (X1，Y1)，(X2，Y2)， ，(X．，Y。)，and to discuss the large sample properties． For this model above，this paper consists ofthree parts，We give respective a kernel shape estimation of re(x)using variable bandwidth regression means， and discuss the large sample properties． Part I We introduce variable bandwidth Nadaraya—Wagon kernel estimator of m(力of complete data,and discuss consistency of the estimator and its derivatives． Part II We present the basic concepts and S011∞propertios of NA sequences，give variable bandwidth Priestley—Chao kernel estimator of complete data,and research rth mean consistency ofthe estimator． PartⅢInthis article．we focus onthe case where sonic y—values ina sample of size”may be missing and all X--values are observed complete， thatis,thedataconsistoftheincompleteobservations(x，，Y。，艿J)，1≤i≤n，6I =1．Based on a set of distribntion free variables under missing response data and fixed design variables under incomplete data,this assumption Was modified Ⅱ further,weighted further,weighted kernel regression estimators are constructed to estimate unknown regressive functions through nonparametric estimation of the missing pattern．All results above are also proven in terms ofdifferent conditiom ofRan -dom elTor variables． Keywords Nonparame=tric regression model；Missing