In the presence of auxiliary information, model-assisted estimators use a working model that links the variable of interest and the auxiliary variables in order to improve the Horvitz-Thompson estimator. The resulting estimators are asymptotically design unbiased and asymptotically more efficient than the Horvitz-Thompson estimator under some regularity conditions and for a wide range of working models. In this work, we adapt model-assisted total estimators to missing at random data building on the idea of nonresponse weighting adjustment. We consider nonresponse as a second phase of the survey and reweight the units in model-assisted estimators using the inverse of estimated response probabilities in order to compensate for the nonrespondents. We develop the asymptotic properties and discuss calibration of the weights of our proposed estimators. We provide formulae for asymptotic variance and variance estimators. We conduct a simulation study that describes the behavior of the proposed estimators.
翻译:在有辅助信息的情况下,模型辅助估算员使用一种工作模型,将利益变量和辅助变量联系起来,以改善Horvitz-Thompson估计值。由此得出的估计值与Horvitz-Thompson估计值相比,在一定的正常条件下和广泛的工作模型中,设计无偏见和无症状效率的简单设计。在这项工作中,我们对模型辅助总估计值进行调整,以适应在不作应答权重调整概念基础上随机数据中缺失的状态。我们认为,不作答复是调查的第二阶段,对模型辅助估计值中的单位进行重新加权,利用估计的反应概率进行反比,以补偿不作答复者。我们开发了无症状特性,并讨论了我们拟议估计值的重量的校准。我们为无应答性差异和差异估计器提供了公式。我们进行了模拟研究,描述了拟议测算员的行为。