...
  • 主题
高级搜索  >
历史搜索 清空历史
首页> 外文期刊>Quality Control and Applied Statistics >A revisit to contingency table and test of independence: Bootstrap is preferred to Chi-square approximations as well as Fisher's exact test
【24h】

A revisit to contingency table and test of independence: Bootstrap is preferred to Chi-square approximations as well as Fisher's exact test

机译：回顾列联表和独立性测试：Bootstrap比卡方近似和Fisher精确测试更可取

获取原文
获取原文并翻译 | 示例
           
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Individuals of a population are classified based on two qualitative attributes, say A and B. A has a classes and B has b classes. Data generated will be in (A_i, B_j) cells with i varying from 1 to a and j varying from 1 to b. The objective is to test the independence of the two variables from the cell frequencies. For small contingency tables, studies have already proposed modified versions of the Chi-square statistic to make the size closer to nominal level (a). However, these test statistics can be used only under restrictive conditions. Fisher's exact test can be used for 2x2 contingency table but will be computationally difficult for larger values of a and b. The test is also considered to be conservative and of lower power. There is no agreement about what is considered a large sample size and what has to be accepted as a small sample size for purpose of applying these tests. Contingency tables with small cell counts are with zero count are called sparse. Zero cell counts can be either due to sampling from non-zero observations are structural zeros with no observations. The approximations from sparse contingency table appear to be poor. The proposed bootstrap versions perform better for sampling zeros and can overcome the above drawbacks. Some studies have proposed methods when the attributes A and B are ordinal, that there exists a natural ordering among A_i values and B_j values. This article assumes that the A and B attributes are nominal. (24 refs.)
机译：人口个体基于两个定性属性（例如A和B）进行分类。A具有一个类，B具有b个类。生成的数据将位于（A_i，B_j）单元中，其中i从1到a变化，j从1到b变化。目的是测试两个变量与小区频率的独立性。对于小的列联表，研究已经提出了卡方统计量的修改版本，以使大小更接近名义水平（a）。但是，这些测试统计信息只能在限制性条件下使用。 Fisher的精确测试可用于2x2列联表，但对于较大的a和b值，计算将很困难。该测试也被认为是保守的且功耗较低。对于应用这些测试的目的，什么被认为是大样本量以及什么被接受为小样本量没有共识。单元格计数少的列联表的计数为零称为稀疏表。零像元计数可能是由于来自非零观测值的采样，也可能是没有观测值的结构零。稀疏列联表的近似值似乎很差。所提出的引导程序版本在对零采样时性能更好，并且可以克服上述缺陷。一些研究提出了当属性A和B为序数时在A_i值和B_j值之间存在自然顺序的方法。本文假定A和B属性是名义上的。 （24参考）

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号