辛普森悖论对描述性统计分析的启示

发布时间：2018-04-01 10:17

本文选题：辛普森悖论　切入点：列联表　出处：《中国统计》2017年08期

【摘要】：正在日常的统计分析实践中,二元列联表是应用最为广泛的统计工具之一。即使是很大型的调查数据,在各类统计软件的辅助下也很容易得到海量的二元交叉列联表;然而如何进行描述性分析并不像表面上看起来那么简单。以辛普森悖论为代表的数据陷阱,一直是个无法回避的重要问题,也是导致二元列联表分析中最容易得出似是而非结论、不可不察之处。我们的探讨也就从这个经典悖论开始。
[Abstract]:In the daily practice of statistical analysis, binary column table is one of the most widely used statistical tools.Even large survey data, with the help of various statistical software, it is easy to obtain a large number of binary cross-column tables; however, how to do descriptive analysis is not as simple as it seems.The data trap represented by Simpson Paradox has always been an important problem that can not be avoided. It is also the most easy to draw specious conclusions in the analysis of binary column tables.Our discussion begins with this classic paradox.
【作者单位】：国家统计局信息景气中心;
【分类号】：C81

【相似文献】