Scarsellone J M
University of Alberta, Rehab Medicine, Edmonton, Canada.
J Speech Lang Hear Res. 1998 Dec;41(6):1341-7. doi: 10.1044/jslhr.4106.1341.
Most research in speech-language pathology relies on observational data collected by human observers or judges. The reliability and generalizability of such measurements are always important considerations. This article reviews classical methods of estimating reliability and proposes that a more powerful approach capable of estimating the dependability of behavioral measurements is available. This approach, based on generalizability theory, provides a practical framework for estimating multiple sources of measurement error in the collection of observational data. Concepts central to generalizability theory are discussed, and a hypothetical data set illustrates the usefulness of generalizability measurements in speech and language research.
大多数言语语言病理学研究依赖于人类观察者或评判者收集的观察数据。此类测量的可靠性和可推广性始终是重要的考量因素。本文回顾了估计可靠性的经典方法，并提出有一种更强大的方法可用于估计行为测量的可靠性。这种基于概化理论的方法为估计观察数据收集过程中的多种测量误差来源提供了一个实用框架。文中讨论了概化理论的核心概念，并通过一个假设数据集说明了概化测量在言语和语言研究中的有用性。
