Menold Natalja, Raykov Tenko
Leibniz Institute for the Social Sciences, Mannheim, Germany.
Michigan State University, East Lansing, MI, USA.
Educ Psychol Meas. 2016 Jun;76(3):454-469. doi: 10.1177/0013164415593602. Epub 2015 Jul 8.
This article examines the possible dependency of composite reliability on presentation format of the elements of a multi-item measuring instrument. Using empirical data and a recent method for interval estimation of group differences in reliability, we demonstrate that the reliability of an instrument need not be the same when polarity of the response options for its individual components differs across administrations of the instrument. Implications for empirical educational, behavioral, and social research are discussed.
本文探讨了复合信度对多项目测量工具各元素呈现形式的可能依赖性。利用实证数据和一种用于信度组间差异区间估计的最新方法，我们证明，当该测量工具各单独成分的反应选项极性在工具的不同施测中存在差异时，其信度不一定相同。文中还讨论了对实证性教育、行为和社会研究的启示。
