Alpha 係數及相關的信度估計方法探討研究

A Note on Coefficient Alpha and Some Related Reliability Estimation Methods

自從 Spearman 於 1904 年發表兩篇日後成為古典測驗理論和因素分析基礎的研究報告之後，古典測驗理論（classical test theory, CTT）逐漸成形，信度的概念和理論也被發展成熟。在眾多信度估計方法中，最負盛名的是 α係數（Cronbach, 1951）。因為運用α係數來報導信度的學者中，未必人人皆都熟悉 α的原理及優劣處，因此本文針對信度的理論與 α係數的原理與計算，做了深入的說明；本文也介紹可以與 α係數一起並用的信度估計方法，然後呈現學者針對這些方法的所進行的比較研究結果，主要結論是對於測驗真實信度的估計，在某些情境下，α的表現不佳，最好在報告測驗資料的α係數時，能同時輔以表現更好的估計值，如 ωₜ 或 λ₄ 等。

 

Since Spearman published two studies in 1904, classical test theory (CTT) had been developed and matured, and the concept of reliability had also been established. Among the methods proposed to estimate the reliability, the coefficient alpha proposed by Cronbach (1951) had emerged to be the most popular one. However, most of those who employed coefficient alpha to estimate the test reliability may not be familiar to the theory foundation. In this paper, the nature and limitation of coefficient alpha had been explicated; and some other reliability estimates, such as,λ₁,λ₂,λ₃,λ₄,λ₅,λ₆,β,ωₜ,ωₕ,and Θ were also introduced. It is suggested that when reporting the reliability for an empirical study, other indices should be presented together with coefficient alpha.