变系数部分线性误差变量模型的岭估计  

Ridge Estimation for Varying-coefficient Partially Linear Error Variable Model

作　　者：曹连英[1] 毕琳 Cao Lianying;Bi Lin(School of Science,Northeast Forestry University,Harbin 150040,China)

机构地区：[1]东北林业大学理学院,哈尔滨150040

出　　处：《统计与决策》2020年第24期25-28,共4页Statistics & Decision

基　　金：中央高校基本科研业务费专项资金资助项目(2572018BC20);黑龙江省自然科学基金资助项目(C201408)。

摘　　要：文章在变量存在测量误差的情况下,对半参数变系数部分线性模型估计进行了研究。为解决协变量之间通常存在的多重共线性问题,给出了参数的岭估计及非参数部分估计表达式,研究其渐近性质,最后进行数值模拟。结果表明:岭估计减小了由复共线性造成的某些参数分量方差异常变大而对非参数估计的影响,当存在复共线性时,岭估计方法优于修正的轮廓最小二乘估计。This paper studies the semiparametric varying-coefficient partially linear models with measure errors in variables.To solve the common multicollinearity problem between covariates,the paper puts forward the ridge estimation of the parameter and the estimate expression for nonparametric part,then studies the asymptotic property,and finally conducts numerical simulation.The results show that ridge estimation reduces the influence of some parameter component variance anomalies caused by multicollinearity on non-parameter estimation,and that the ridge estimation method is superior to the modified contour least squares estimation when multicollinearity exists.

关 键 词：变系数误差变量模型 轮廓最小二乘法 岭估计 

分 类 号：O21[理学—概率论与数理统计]