Nearest Neighbor Imputation for Survey Data
Nearest Neighbor Imputation for Survey Data
复制标题
DOI：
--
复制
发表时间：
2000
影响因子：
1.1
通讯作者：
Jiahua Chen;J. Shao
中科院分区：
文献类型：
--
作者：
Jiahua Chen;J. Shao
在抽样调查中，归罪通常被用来补偿无反应(Kalton 1981；Sedransk 1985；Rubin 1987)。最近邻法(NNI)在加拿大统计局、美国劳工统计局和美国人口普查局进行的许多调查中被使用，而且这种趋势将继续下去，因为有了计算机软件通用编辑和归罪系统，它提供了一种执行NNI的简单方法(Cotton 1991；Rancourt，Saérndal和Lee 1994；Kovar，Whitbridge和MacMillan 1998)。让我们首先介绍最简单情况下的NNI方法。考虑一个二元样本(x1；y1)，.。。；xn；Yn)，并且假设观察到了n个y值中的r个(受访者)，其余的m个n？r y值缺失(非受访者)，并且观察到了所有的x值。为简明起见，我们假定YR_1；。。。；YN失踪。NNI方法通过yi来推算缺失的yj，r i 1#j#n，其中1#i#r和i是由x变量测量的j的最近邻居，即，i满足
Imputation is commonly applied to compensate for nonresponse in sample surveys (Kalton 1981; Sedransk 1985; Rubin 1987). The nearest neighbor imputation (NNI) method is used in many surveys conducted at Statistics Canada, the U.S. Bureau of Labor Statistics, and the U.S. Census Bureau, and this trend will continue because of the availability of a computer software, the Generalized Edit and Imputation System, which provides a simple way of performing NNI (Cotton 1991; Rancourt, SaÈrndal, and Lee 1994; Kovar, Whitridge, and MacMillan 1998). Let us begin with an introduction of the NNI method in the simplest case. Consider a bivariate sample (x1; y1),. . . ; xn; yn) and suppose that r of the n y-values are observed (respondents), the rest of m n ÿ r y-values are missing (nonrespondents), and all x-values are observed. For simplicity we assume that yr1; . . . ; yn are missing. The NNI method imputes a missing yj, r 1 # j # n, by yi, where 1 # i # r and i is the nearest neighbor of j measured by the x-variable, i.e., i satis®es