协变量缺失下基于诱导光滑方法的加权分位数回归

在部分协变量随机缺失机制下的分位数回归模型中,提出回归参数的诱导光滑加权估计及其渐近协方差估计,证明了诱导光滑加权估计和经验似然加权估计有相同的渐近协方差,且诱导光滑加权估计的渐近协方差估计也是相合的,并给出了诱导光滑加权估计及其渐近协方差估计的高效算法.模拟结果表明,新方法在有限样本下表现优良.

Weighted Quantile Regression Based on InducedSmoothing Method with Missing Covariates

We proposed an induced smoothing based weighted estimator of regression parameter and an estimator of its asymptotic covariance in quantile regression model of partial covariates with random missing mechanism. We showed that the induced smoothing based weighted estimator and empirical likelihood based weighted estimator had the same asymptotic covariance and the estimator of the asymptotic covariance of the induced smoothing based weighted estimator was also consistent. We gave an efficient algorithm for calculating the induced smoothing based weighted estimator of regression parameter and the estimator of its asymptotic covariance. Simulation results show that the proposed method performs well in finite samples.