ESTIMATION WITH NEAREST NEIGHBOUR IMPUTATION AT STATISTICS CANADA
ESTIMATION WITH NEAREST NEIGHBOUR IMPUTATION AT STATISTICS CANADA
复制标题
DOI：
--
复制
发表时间：
2002
期刊：
影响因子：
--
通讯作者：
E. Rancourt
中科院分区：
文献类型：
--
作者：
E. Rancourt
在进行调查时，人们总是要面对数据缺失的问题。几十年来，调查统计学家一直使用捐赠者归罪技术来治疗无应答。人们只需想一想，当丢失的记录被受访者的卡片替换时，穿孔卡片的时间。HotDeck(捐赠者的一种形式)就是这样诞生的。多年来，捐赠者归责经历了几次改进，主要的一次是利用了受访者和非受访者都可以使用的辅助变量。在这种情况下，在试图找到与丢失的记录最相似的供体时，使用最近的记录(根据某种距离度量)。自Sande(1981)以来，这种最接近匹配的热甲板一直被称为最近邻归责。然而，特征，但最近邻归属具有使用辅助信息的优势。它也有一种直观的吸引力。这些原因，再加上Chen和Shao(1997)和第4节中讨论的良好的形式性质，现在使最近邻归罪成为任何归罪策略的主要候选者。
In conducting surveys, people have always had to face problems of missing data. For a few decades, survey statisticians have made use of donor imputation techniques to treat nonresponse. One only has to think of the time of punch cards when a missing record was replaced by the card of a respondent. That is how hotdeck (a form of donor) imputation was born. Over the years, donor imputation went through several refinements, the major one being making use of auxiliary variables available for both respondents and nonrespondents. In this case, in an attempt to find a donor most similar to the missing record, the closest record (according to some distance measure) is used. This closest matching hot-deck has been called nearest neighbour imputation since Sande (1981). feature however, but nearest neighbour imputation has the advantage of using auxiliary information. It also has an intuitive appeal. These reasons, plus the good formal properties discussed in Chen and Shao (1997) and in Section 4 now make nearest neighbour imputation a prime candidate for any imputation strategy.