nonparameter 2015 (Student)课件.pptVIP

注意报告的顺寻：先描述，再推断；先方差分析、再多重比较。 数据类型和统计学描述 (numerical data, normal distribution，指标、小数位数、单位) 设计类型和分析方法–方差分析条件 (completely random design, two conditions) 在做多重比较时，也需要建立假设检验 多重比较方法的选择也要有依据 表格（描述、方差分析）要规范：标题、表格线；选择性的报告SPSS中的结果。 即使对于双侧检验，最终下结论时，遇到P0.05，也要根据样本提供的信息将双侧结论“落实”为单侧结论。如：本题中得出春夏处于同一水平，秋冬季处于另一水平，又因为春夏的样本均数高于秋冬季，则最终认为春夏处于同一水平，且高于秋冬（另一水平）。 Lesson 6 Non-parametric test 2015 Objectives: 1. To understand the concept and the advantages of nonparametric tests based on ranks 2. To master the methods to rank . 2. Factors to determine power. 3. To master the Nonparametric Tests procedures in SPSS. Review The way to draw a conclusion according to t-value in t-test. The way to draw a conclusion according to T-value in Wilcoxon’s signed rank sum test. -3 -2 -1 0 1 2 3 t分布图 P=0.025 P=0.025 x t0.05,ν （双侧） 对子 差值 T+ T- T+ T- 3.8 4.6 3.0 4.1 6.0 3.5 8.7 3.5 5.8 4.7 -0.8 -1.1 2.5 -2.4 4.0 Discussion:1. Which one will you choose to compare two samples, t-test or rank sum test? What are the reasons? 2. If the conditions of t-test are satisfied, compared with t-test for two samples, rank sum test: A More likely to reject H0 B Less likely to reject H0 C The results are the same D The results are different E None of the above 3. When we use rank sum test to compare two samples: A The same sample sizes are required B Rank all the observations one by one orderly and give a mean rank to the same value C Rank all the observations one by one orderly and omit zero D It’s a parametric test if the statistic Z is calculated when the sample sizes are large enough E None of the above 4. What are proper ways to improve power? A Increase sample size. B Increase variation in populations. C Increase the difference between the population means. D Take a lower risk for type I error. E Take a lower risk for type II error. 5. Two medicines were applied to decrease triglyceride (甘油三酯) in serum. 189 patients with high TG were allocated int