護理研究中常遇到的統計問題實例探討

An Empirical Discussion of Commonly Encountered Statistical Problems in Nursing Research

盧成皆(Sing-Kai Lo)

《護理研究》 4卷3期 (1996/09) Pp. 298-306

摘要

統計學之應用於護理研究中越來越多、越來越深入，但有些統計問題並非直接從課本、論文或軟體中便可找到答案，而是需要使用者融會貫通、靈活運用才可。本文即針對三個常遇到的問題以實例加以探討及說明。其一爲在使用問卷作爲研究工具之前，基於研究的架構與目的，問卷的再測信度很多時候都需要建立，但若以皮爾森氏相關係作爲信度的指標，會有很多頗爲嚴重的缺點，故介紹使用內在等級相關係數作爲指標，並說明基優點。此外，樣本數目的的計算幾乎是每個護理研究所面對的問題，現在已經有一些統計軟體可供使用，唯這些軟體的使用，需仰賴使用者鍵入母體的一些參數或其估值方可，在缺乏這些參數估值時，本文建議使用一種漸進式的計算方法以解決問題。迴歸法的使用在護理研究中非常普遍，但若在簡單線性迴歸及複迴歸中都沒有統計意義的變項是否就可以忽略呢？本文即以交互作用的存在爲例，說明這些變項亦可能有其解釋能力，研究者宜小心處理。

關鍵字

統計方法 ； 再測信度 ； 樣本數目 ； 迴歸分析

並列摘要

Statisical methods are now more widely and thoroughly used in nursing research. However, solutions for some commonly encountered statistical problems cannot be found directly from textbooks or journal articles. Sometimes results can only be obtained through indirect calculations using figures available from outputs. In this paper, we discuss the drawbacks of the traditionally used Pearson correlation in test-retest reliability assessment, and suggest the use of different types of intra-class correlations. On the other hand, population parameters or their estimates are required for sample size calculations. Very often these parameters are unknown. Under such circumstances, we introduce the use of a sequential method as an option. Regression analysis is one of the most commonly used techniques in nursing research. For variables not significant in either simple or multiple linear regression, researchers almost always conclude that these variables do not have predictive power. Yet, these variables may still confound or interact with each other to change the interpretations of the final model. Data analysts therefore have to be extremely careful in model building using regression techniques.

並列關鍵字

statistical methods ； test-retest-reliability ； sample size ； regression analysis