護理研究者經常使用Cronbachl's alpha(α)係數評估量表的內在一致性,通常是先設定一個可以接受的範圍(如:0.7至0.9;亦有建議0.3至0.7),若係數過低或過高,一般便認為量表的內在一致性不足,或題目與題目間相關性過強,都需要修改問卷,甚至刪除部份題目;但若結果滿意,則會覺得把題目的分數相加成一總分是合理的,可作後續統計分析之用。本文即以此為出發點,設計了一系列的假設性數據,利用統計報表,以一般護理研究者容易了解的文字及方式,說明 Cronbachl's α係數的計算過程,從而解釋當。係數偏低時,除量表的修辭或有關問題外,亦需要考慮樣本的收集方法是否適宜。另一方面,當α係數介乎於0.7至0.9(或0.3至0.7)之間時,卻又不一定絕對代表量表之內在一致性沒有問題。綜言之,分析時應仔細思考,臨床與統計並重,不宜過份仰賴單一係數。
Cronbach's alpha (α) is a commonly used measure of internal consistency in nursing research. Typically, if the sample a coefficient is not within a pre-specified range (e.g. 0.7 to 0.9; 0.3 to 0.7 has also been suggested), then it will be concluded that the research tool is not sufficiently consistent internally, or some questions within the same construct are too strongly correlated. In either case, the questionnaire will have to be reviewed, sometimes with a few questions deleted. But if satisfactory results are obtained, then a total score will be constructed from items belonging to the same construct for subsequent statistical analysis. However, using a series of hypothetical data, we show in layman's terms that a low a score does not necessarily imply there are problems with the wordings or the questions themselves. The sampling procedure may have to be revised instead. On the other hand, even when the a coefficient lies within the pre - defined range, the reliability and hence validity of the questionnaire can still be doubtful. We therefore strongly advise researchers not to rely too heavily on a single measure. A more thorough consideration from both the clinical and statistical points of view is essential.