摘要
The existence of non response damages the precision of estimators in survey severely. The common countermeasure is imputation and weighting, the former makes use of the auxiliary information, the latter estimates by response rate. Each of them has merits as well as weakness. In order to incorporate the merits of the methods mentioned above, we put forward calibration estimation, which suggests adjusting the preliminary weights by auxiliary information at the stage of estimating. Marke the best of the relations between the independent variables and the dependent variable, use appropriate estimation method, and you’ll get a good estimator for the sum of the target variable.
The existence of non response damages the precision of estimators in survey severely. The common countermeasure is imputation and weighting, the former makes use of the auxiliary information, the latter estimates by response rate. Each of them has merits as well as weakness. In order to incorporate the merits of the methods mentioned above, we put forward calibration estimation, which suggests adjusting the preliminary weights by auxiliary information at the stage of estimating. Marke the best of the relations between the independent variables and the dependent variable, use appropriate estimation method, and you'll get a good estimator for the sum of the target variable.
出处
《统计研究》
CSSCI
北大核心
2002年第6期32-35,共4页
Statistical Research
