第六堂(01-08)-慈济医院.ppt

* * * * * * * * * * * * * * * * * * * * * * * * * * Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile . Copyright 2004-2011 Aspose Pty Ltd. Type I error 問題 進行研究時，若有多組比較時，應事前即定義好組別 不應為了產生統計意義而重新分組 分組應符合常理，不然應於方法中詳細敘述 若結果為連續變項，多組比較時，應採ANOVA 及 post-hoc (Bonferonni or Tukey, etc.)，不可兩兩相比，產生過多之type I error. Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile . Copyright 2004-2011 Aspose Pty Ltd. Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile . Copyright 2004-2011 Aspose Pty Ltd. Example: 若有10組學生比較身高，要用何種檢定呢? ? One-way ANOVA及post-hoc 若兩兩檢定會有什麼問題嗎? ?type I error 可能會達到 45x0.05 解決方法: Bonferroni Multiple comparisons procedure, 此時達統計意義之標準為 0.05/45=0.0011 Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile . Copyright 2004-2011 Aspose Pty Ltd. Example 若有甲乙兩班各5人比較體重高低，要用何種檢定? Wilcoxon Rank-Sum test or Mann-Whitney U test 因為個案數過少，採用無母數方法 Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile . Copyright 2004-2011 Aspose Pty Ltd. 使用Chi-square test常見錯誤 應先建立要檢測之虛無假說(null hypothesis) ，再選定檢定變項。 若列聯表中有期望值5時，應改用Fisher’s exact test 評估結果 若組別過多，可考慮併組 Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile . Copyright 2004-2011 Aspose Pty Ltd. Tx * Result Crosstabulation Tx 1.00 2.00 Total Count Expected Count Count Expected Count Count Expected Count Result 1.00 8 6.2 1 2.8 9 9.0 2.00 10 11.8 7 5.2 17 17.0 Total 18 18.0 8 8.0 26 26.0 Chi-Square Tests Pearson Chi-Square Continuity Correctionb Likelihood Ratio Fishers Exact Test Linear-by-Linear Association N of Valid Cases a. 1 cells (25.0%) have expected count less than 5. The minimum expected count is 2.77. b. Computed only for a 2x2 table Value 2.497a 1.285 2.783 2.401 26 df 1 1 1 1 Asymp. Sig. (2-sided) .114 .257 .095 .121 Exact Sig. (2-sided) .190 Exact Sig. (1-sided) .128 Evaluation only. Created with Aspose.Slides for .NET 3.5 Client Profile . Copyright 2004-2011 Aspose Pty Ltd. 2.3 統計方法陳述不完整 雙尾或單尾檢定 Two