r×c 行列表卡方值分割的概念及运用详解.pdfVIP

R×C 行列表卡方值分割的概念及运用 The Partitioning of Chi-Square Statistics from R×C Contingency Tables ： Concept and Application 万文胜 蔡倩方 [摘要]在临床医学研究中研究人员经常要对 R×C 行列表作统计分析。当2 检验有统计意义时，就有必 要作进一步的多重比较，特别是作2 分割以分析具体什么亚效应导致了原表的统计意义。本文阐述了 分割 Pearson 2 2 2  和 likelihood ratio  的统计原理和实践中关于 分割的常见错误观点。本文分别列举 了分割 likelihood ratio 2 和 Pearson 2 所采取的两类不同的统计方法：1）采用类似于析因分析中正交 对比的概念，根据行列正交对比基将原行列表分割成(r-1)×(c-1)个相互独立的四格表，然后分别计算自 由度为 1 的 likelihood ratio 2 n 2  值。2 ）分割 Pearso  时，则先根据行列正交对比基对原行列变量分别 1 进行正交转换，然后计算新的正交行列变量两两之间的 Pearson 相关系数，最后导出自由度为 1 的 Pea 2 现。 rson  。两种分割方法均用血型和癌症的例子加以说明同时均容易在常用统计软件中实 1 关键词：正交对比，卡方分割，行列表 作者单位：518001 广州，广州盈哲科技（万文胜，vincent@toeach.net ） ，广东省妇幼保健医院（蔡倩 方） [Abstract]Statistical analysis of R×C contingency table is commonly performed in clinical studies. Most often in case of having significant p-values for Chi-square tests it is necessary to do sub-analysis in terms of multiple comparisons or a specific type of multiple comparisons- Partition of Chi-square statistics from the original contingency tables. This paper describes the statistical theories behind the partitioning of likelihood ratio and Pearson Chi-square statistics from contingency tables while cross-examining some common misconceptions associated with the partition. Two different but related methodologies to partition the likelihood ratio and Pearson Chi-square statistics from contingency tables are presented in the paper separately: 1) the orthogonal contrasts, the same concept used in Analysis of Var