缺失数据下Jackknife方差估计量的渐近设计无偏性

ASYMPTOTIC DESIGN-UNBIASEDNESS OF JACKKNIFE VARIANCE ESTIMATORS IN THE PRESENCE OF MISSING DATA

在实际的调查中经常会出现缺失数据．如何处理这种情况下总体目标量的估计是一个重要问题．Zou等(2002)对缺失数据情况下的样本轮换方法证明了他们所提出的线性化．Jackknife方差估计量在均匀回答下是估计量方差的近似的设计无偏估计．这一性质对于．Jackknife方差估计量的使用提供了重要依据．对于其它情况下．Jackknife方差估计量是否也具有这一性质无疑是一个有意义的问题．作者旨在研究文献中已提出的若干．Jackknife方差估计量的渐近设计无偏性．我们的结果表明Zou等(2002)所注意到的Jackknife方差估计量的渐近设计无偏性具有一定的普遍性。

In practical surveys, non-response is often unavoidable. How to estimate the population quantities of interest is an important problem in this case. Recently, Zou et al. （2002） have found that for the sample rotation method with missing data, Jackknife variance estimators they proposed are asymptotically design-unbiased under uniform response. This properly provides a solid basis to use Jackknife variance estimators. This paper aims at studying the similar property of some other Jackknife variance estimators existing in the literature. Our results show that the asymptotic design-unbiasedness for Jackknife variance estimators universally holds.