Modeling Ambiguity in Census Microdata Allocations to Improve Demographic Small Area Estimates
Modeling Ambiguity in Census Microdata Allocations to Improve Demographic Small Area Estimates
复制标题
DOI：
10.1111/j.1467-9671.2012.01366.x
复制
发表时间：
2013-06-01
影响因子：
2.4
通讯作者：
Nagle, Nicholas N.
中科院分区：
文献类型：
--
作者：
Leyk, Stefan;Buttenfield, Barbara P.;Nagle, Nicholas N.
本文描述了在存在潜在歧义的情况下，将人口统计微观数据分配到诸如人口普查区等小型枚举区域的方法。最大熵方法估算的人口权重受到约束，以匹配一组人口普查区域级汇总统计数据。一旦分配，对家庭特征进行总结，以修订区域一级人口汇总统计的估计，并得出模糊性的衡量标准。在预期变化的背景下，将修订后的摘要统计数据与原始摘要进行比较。分配歧义被量化为每个家庭在所有人口普查区的估算样本权重分布的函数，并通过计算混乱和分配到任何人口普查区的多样性的度量。这里报告的过程允许在分配决策中固有的模糊性方面对家庭进行区分。模糊性评估是迄今为止空间分配工作中被忽视的一个重要组成部分，但对于人口统计学家和小面积估算的用户来说，它可以被视为重要的附加知识。对于大多数被测试的变量，修订后的区域水平总结与原始区域总结统计高度相关。除了对个别家庭的评估外，还可以计算个别地区的平均分配不确定性，并将其与分配过程中未使用的人口特征联系起来。
This article describes a methodology for allocating demographic microdata to small enumeration areas such as census tracts, in the presence of underlying ambiguities. Maximum Entropy methods impute population weights that are constrained to match a set of census tract-level summary statistics. Once allocated, the household characteristics are summarized to revise estimates of tract-level demographic summary statistics, and to derive measures of ambiguity. The revised summary statistics are compared with original tract summaries within a context of expected variation. Allocation ambiguity is quantified for each household as a function of the distribution of imputed sample weights over all census tracts, and by computed metrics of confusion and variety of allocation to any census tract. The process reported here allows differentiation of households with regard to inherent ambiguity in the allocation decision. Ambiguity assessment represents an important component that has been neglected in spatial allocation work to date but can be seen as important additional knowledge for demographers and users of small area estimates. For the majority of tested variables, the revised tract level summaries correlate highly with original tract summary statistics. In addition to assessments for individual households, it is also possible to compute average allocation ambiguity for individual tracts, and to associate this with demographic characteristics not utilized in the allocation process.