Bartko J J
Division of Epidemiology and Services Research, National Institute of Mental Health, Bethesda, Maryland 20892.
Stat Med. 1994;13(5-7):737-45. doi: 10.1002/sim.4780130534.
An assessment of measurement agreement made by devices, laboratories, or raters is important in medical practice and research. The setting in which each randomly selected subject is rated by the same two raters raises assorted questions regarding rate agreement. The intraclass correlation coefficient (ICC) is one measure of reliability. The paired t-test can be used to evaluate the overall ratings or bias of the two raters, while their variances can be assessed with Pitman's test. The Bradley-Blackwood test can be used for a simultaneous test of their means and variances. A single method that provides results for these approaches is proposed and the bivariate confidence ellipse is suggested to provide boundaries for dispersion.
对设备、实验室或评估者所达成的测量一致性进行评估在医学实践和研究中很重要。由相同的两名评估者对每个随机选择的受试者进行评分的情况引发了关于评分一致性的各种问题。组内相关系数（ICC）是可靠性的一种度量。配对t检验可用于评估两名评估者的总体评分或偏差，而他们的方差可用皮特曼检验进行评估。布拉德利 - 布莱克伍德检验可用于同时检验他们的均值和方差。本文提出了一种能为这些方法提供结果的单一方法，并建议使用双变量置信椭圆来提供离散度的边界。
