北京大学学报(自然科学版)

一类重要辅助信息在一类因果模型中可识别的充要条件

童行伟1, 郑忠国1, 张艳艳2, 崔现伟1   

  1. 1北京大学数学科学学院,北京,100871;2天津师范大学,天津,301700
  • 收稿日期:2001-06-30 出版日期:2002-09-20 发布日期:2002-09-20

Sufficiency and Necessity for the Identifiability of An Important Type of Ancillary Information in A Causal Model

TONG Xingwei1, ZHENG Zhongguo1, ZHANG Yanyan2, CUI Xianwei1   

  1. 1School of Mathematical Sciences, Peking University, Beijing, 100871; 2Tianjin Normal University, Tianjin, 301700
  • Received:2001-06-30 Online:2002-09-20 Published:2002-09-20

摘要: 在流行病学和社会科学中,大量的研究是估计因果效应。一般而言,没有任何假设的条件下,因果效应无法获得,即不给因果变量任何辅助信息,因果效应无法识别。作者提出了辅助信息可识别的定义,并且针对一类比较普遍且重要的辅助信息得出它在一类因果模型中对4种因果效应可识别的充要条件,最后将这结果推广到一般情况。

关键词: 可识别, 因果效应, 辅助信息

Abstract: Much of epidemiological and social science research is devoted to estimation of causal effect. Generally, the causal effect is not identifiable without any constraint which is called the ancillary information in this paper. First the definition of the identifiability for the ancillary information is proposed and then sufficient and necessary conditions for the identifiability of an important type of ancillary information are discussed. Finally, the results are expanded to a more common case.

Key words: identifiability, causal effect, ancillary information

中图分类号: