信度評估之修訂與校正及其在管理實證研究上之應用

Abstract

本研究計畫為兩年期計畫，擬分兩部分採兩階段分年進行，第一部分為信度評估之修訂 與校正研究，第二部分運用第一部分研究結果進行管理實證分析。複合信度(composite reliability, 記為CR)與α 係數常用來量測一構念各指標問項之內部一致性，0.7 為普遍 之接受門檻，由母體的角度觀之，α 係數為 CR 之下限，故以0.7 作為樣本CR 之接受 門檻有偏低之虞，本計畫第一部分擬推導出α 與 CR 之差異公式D，據此即可訂定樣 本CR 之合理接受門檻。然而，在問項間相關係數平均值固定不變下，α 係數會隨問項 數之增加而提高，此一特性(問項數多時會導致信度高估)雖眾所周知，但在文獻上似未 有具體的應對之道，故本研究擬接著就此問題進行深入探討，並提出一「校正後α 係 數」以反映真實之內部一致性(不受問項數多寡影響)。以上第一部分(於第一年進行)有 關信度量測探討所得結果將提供給第二部分(於第二年進行)之實證研究運用，實證研究 有兩個議題，一為「組織公民行為與人格特質、程序公平氣候、群體凝聚力及顧客結果 關聯之多層次分析」；另一為「背景音樂如何發揮網路購物效益？」，將分別獨立進行， 擬定研究架構，檢驗研究假設，討論管理意涵。

Coefficient alpha is a reliability index to assess internal consistency of the items used to measure the same construct. It is well known that, for a fixed average correlation coefficient, alpha will inflate when the number of items increases. A corrected coefficient alpha will be proposed in this study to reflect actual internal consistency, not influenced by the number of items. On the other hand, composite reliability is always used when CFA is conducted. The difference between coefficient alpha and composite reliability will be specifically addressed in this study, based on which the reliability can be assessed in a more objective way. The results obtained will be used to assess reliabilities for two empirical studies. One is 「the relationships among organizational citizenship behavior and personalities, procedural justice climate, group cohesiveness, and customer outcomes」, and the other is 「how could online background music stimuli work?」. For each study, we will propose research framework, test research hypotheses, and discuss managerial implications.