Massey J T, Hoffman K L
National Center for Health Statistics, Centers for Disease Control, Hyattesville, MD 20782.
Stat Med. 1989 Mar;8(3):367-77. doi: 10.1002/sim.4780080318.
Since its creation in 1960, the National Center for Health Statistics (NCHS) has placed a high value on the quality of the information and statistics collected and published by its data systems. An important component of a comprehensive data quality monitoring system for a statistical agency such as NCHS is the comparison of statistics between data systems. Between-data-systems comparisons are used to monitor the consistency and comparability of statistics across data systems and between NCHS data systems and outside sources of data. This paper focuses on between-data-systems comparisons and describes a number of methodological analyses that can be used to evaluate data quality. The methodological analyses presented include the evaluation of time trends, the estimation of survey method effects, the evaluation of response error, the measurement of definitional and concept effects, and the detection of inconsistencies between data systems. The paper concludes with a discussion of the complexities of the between-data-systems comparisons and an assessment of the net benefits from such comparisons.
自1960年成立以来，美国国家卫生统计中心（NCHS）一直高度重视其数据系统所收集和发布的信息及统计数据的质量。对于像NCHS这样的统计机构而言，全面的数据质量监测系统的一个重要组成部分是数据系统之间统计数据的比较。数据系统间的比较用于监测各数据系统之间以及NCHS数据系统与外部数据源之间统计数据的一致性和可比性。本文重点关注数据系统间的比较，并描述了一些可用于评估数据质量的方法分析。所介绍的方法分析包括时间趋势评估、调查方法效应估计、应答误差评估、定义和概念效应测量以及数据系统间不一致性检测。本文最后讨论了数据系统间比较的复杂性，并评估了此类比较的净收益。
