a propensity score adjustment method for regression models with nonignorable missing covariates：一个具有不可忽略缺失的协变量回归模型校正倾向评分方法.pdfVIP

Computational Statistics and Data Analysis 94 (2016) 98–119 Contents lists available at ScienceDirect Computational Statistics and Data Analysis journal homepage: /locate/csda A propensity score adjustment method for regression models with nonignorable missing covariates Depeng Jiang a,∗, Puying Zhao a , Niansheng Tang b a Department of Community Health Sciences, University of Manitoba, Canada b Department of Statistics, Yunnan University, PR China a r t i c l e i n f o a b s t r a c t Article history: In a linear regression model with nonignorable missing covariates, non-normal errors or Received 7 July 2014 outliers can lead to badly biased and misleading results with standard parameter estima- Received in revised form 21 July 2015 tion methods built on either least squares- or likelihood-based methods. A propensity score Accepted 21 July 2015 method with a robust and efficient regression procedure called composite quantile regres- Available online 14 August 2015 sion for parameter estimation of the linear regression model with nonignorable missing covariates is proposed. Semiparametric estimation of the propensity score is based on the Keywords: exponentially tilted likelihood approach. Asymptotic properties of the proposed estimators Exponentially tilted likelihood Composite quantile regression are systematically investigated. The proposed method is resistant to heavy-tailed errors or Not missing at random outliers in the resp