態度量表的心理計量學分析：2003年TEDS統獨態度量表的研究

Abstract

政治學者在進行實證分析時，一般都過度輕忽理解測量理論的重要性，導致在測量上有許多問題發生而不自覺，這些問題不見得每一次都會有嚴重的影響，可是一旦發生將從根本上威脅到研究的可信度，使得一篇可能是很好的實證研究失去它原有的價值。本文透過對測量理論原理和核心議題的探討，來強調具備測量理論基礎知識的重要性，並針對五項基本假設，來對於十六種不同統獨量表進行心理計量學的分析。研究發現，任何一項假設都可能左右迴歸分析的結果，甚至在相關係數高達0.995的情況下也是如此，這顯示政治學者在進行測量工作時，應慎重面對測量問題可能對研究成果在科學價值上的減損。

Political scientists usually pay little attention to the understanding of measurement theories when they conduct a quantitative empirical analysis. This engenders many problems beyond their knowledge. While not every problem has a substantial impact on the research, it may fundamentally threaten the credibility and nullify the achievement once it happens. This article tries to argue the importance of proper understanding on some basic knowledge of measurement theories through a discussion of the principles and nuclear issues. Furthermore, a psychometric analysis of sixteenth different scales, based on five psychometric assumptions, is carried out for the unification-independence attitude. The result indicates each of the five assumptions is possibly influential to the regression analysis, even if the simple correlation of the two scales is as high as 0.995. This finding leads to an important conclusion: political scientists should have a serious discussion about the assumptions of their measurement and evaluate the possible impact which may discredit the scientific value of their research.