Haas M
Research Division, Western States Chiropractic College, Portland, OR 97230, USA.
J Manipulative Physiol Ther. 1995 Jan;18(1):10-5.
Examiner reliability is often investigated to make generalizations about a profession's performance with various diagnostic tests. Although the evaluation of interexaminer reliability is straightforward, assessment of intraexaminer reliability can be problematic. Estimations of intrarater reliability can be inflated due to correlated error and the difficulty in blinding the examiners. A statistical method is presented that permits the investigator to compute intraexaminer as well as interexaminer reliability and precision from the findings of an interexaminer reliability investigation. In this approach, intraclass correlation coefficients are constructed from variance components estimated from a simple repeated-measures design: each of two or more raters evaluates each subject one time. The method can be applied to nominal and ordinal as well as interval data. Not only can spurious results be avoided, but time and funds can be saved by reducing the required number of subject ratings.
通常会对检查者的可靠性进行调查，以便通过各种诊断测试对某一职业的表现做出概括。虽然检查者间可靠性的评估很简单，但检查者内可靠性的评估可能会有问题。由于相关误差以及对检查者进行盲法处理的困难，检查者内可靠性的估计可能会被夸大。本文提出了一种统计方法，使研究者能够根据检查者间可靠性调查的结果计算检查者内以及检查者间的可靠性和精密度。在这种方法中，组内相关系数是根据从简单重复测量设计中估计的方差成分构建的：两个或更多的评分者每人对每个受试者评估一次。该方法可应用于名义数据、有序数据以及区间数据。不仅可以避免虚假结果，而且通过减少所需的受试者评分数量还可以节省时间和资金。
