如何選擇合適雙樣本的統計檢定方法

How to Choose Appropriate Two-Sample Statistical Tests

臨床研究者閱讀醫學論文時，對於一些統計檢定方法可能會不知其原理及其適用情境，又在資料收集完成後，很可能不知道如何選擇合適統計檢定方法的困擾。本文利用－表格來綜合常用的雙樣本分析方法，然後藉一個模擬的臨床試驗範例說明如何選擇一個統計檢定方法。除了「統計導向」的考量外，尚有「結果導向」、「慣性導向」、「一致導向」等選擇考量。幫助讀者瞭解在分析數據時，除了熟悉統計方法的適用時機外，尚有其他選擇考量。在呈現結果時，應以客觀及科學態度作審慎的選擇。閱讀研究論文時，應注意研究者有可能是依「結果導向」或「慣性導向」來呈現結果。

When reading journal papers, researchers may face a problem of not under-standing the principle or rationale of statistical tests. When performing data analysis, researchers do not know how to choose appropriate statistical tests.The aims of this study were to: (1) summarize 2-sample statistical tests in a table so that researchers can easily refer; (2) simulate a clinical trial data and illustrate how to choose statistical tests. Other than ”statistical” principle, ”outcome-oriented” reason, ”tradition-oriented” reason and ”consistency” were involved in the derision of choosing statistical rests when addressing a research question. We suggest that researchers present their results should be based on an objective and scientific perspective. When reading a journal paper, one should realize that the results presented may come from the decision of ”outcome-oriented” or ”tradition-oriented” reason.