响应变量缺失下半参数变系数EV模型的约束统计推断

Constrained Statistical Inference for Semi-parametric Variable Coefficient EV Model With Missing Response Variables

文章研究了半参数变系数EV模型在线性约束条件下的估计和检验问题,当响应变量缺失、非参数部分协变量带有测量误差时,利用局部纠偏的Profile最小二乘估计、Lagrange乘子方法和借补技术构造了回归模型参数分量两类纠偏约束估计量。此外,为了检验线性约束条件,构造了借补的Profile Lagrange乘子检验统计量,并通过蒙特卡洛数值模拟验证估计量和检验统计量的有效性。

This paper studies the estimation and test of semi-parametric variable coefficient EV model under linear constraint. When the response variable is missing and the non-parametric part of the covariable has measurement error, two types of bias-corrected constrained estimators of parameter component in regression model are constructed based on the locally corrected Profile least square estimation, the Lagrange multiplier method and the imputation technology. In addition, the interpolated Profile Lagrange multiplier test statistic is also constructed to examine the linear constraints, and the validity of the estimators and the test statistic are verified by Monte Carlo simulation.