A General View of Estimation for Two Phases of Selection with Applications to Two-Phase Sampling and Nonresponse
本文研究两阶段抽样和无应答情况下的估计问题，给出两阶段抽样的一般结果，重点讨论回归估计和方差估计，并将结果应用于无应答场景，发现回归型估计在响应机制错误时比单纯加权更稳健。
Summary This paper is meant as a contribution to estimation (a) in the case of two-phase sampling, (b) in the case of nonresponse. In both cases, there is subsample selection. The difference is that the subsampling obeys a known distribution in two-phase sampling, an unknown distribution ('the response mechanism') in the nonresponse case. General results are given for two-phase sampling, with emphasis on regression estimation and on the problem of variance estimation. The results, with necessary modifications, are applied to the nonresponse situation. In particular, if the assumed response mechanism is false, we find that regression-type estimators give more robust confidence statements than if weighting alone is employed.