通过重参数化有向无环图估计序贯处理效应
统计方法学
2013-02-08 v1
摘要
参数化有向无环图所表示分布的标准方法是为每个随机变量给定其父节点的条件分布插入一个参数族。我们表明,当目标是检验或估计序贯施加处理的效应时,这种自然参数化存在严重缺陷。通过使用结构嵌套模型 (structural nested models) 对图进行重参数化,可以避免这些缺陷。
引用
@article{arxiv.1302.1566,
title = {Estimation of Effects of Sequential Treatments by Reparameterizing Directed Acyclic Graphs},
author = {James M. Robins and Larry A. Wasserman},
journal= {arXiv preprint arXiv:1302.1566},
year = {2013}
}
备注
Appears in Proceedings of the Thirteenth Conference on Uncertainty in Artificial Intelligence (UAI1997)