Measuring Model Misspecification: Application to Propensity Score Methods with Complex Survey Data.
Measuring Model Misspecification: Application to Propensity Score Methods with Complex Survey Data.
复制标题
DOI：
10.1016/j.csda.2018.05.003
复制
发表时间：
2018-12
影响因子：
1.8
通讯作者：
Stuart EA
中科院分区：
文献类型：
--
作者：
Lenis D;Ackerman B;Stuart EA
模型错误指定对于任何基于参数模型的分析都是一个潜在的问题。然而，在因果推理的背景下，模型错误说明的度量和后果还没有得到很好的形式化。提出了一种模型错误说明的度量方法，并研究了非实验因果推理方法中模型错误说明的后果。然后，该指标被用来探索哪些估计者对结果和/或治疗分配模型的错误指定更敏感。我们考虑了三个常用的治疗效果估计器，它们都依赖于倾向评分：(1)完全匹配，(2)1：1近邻匹配，(3)加权。在两种不同的抽样设计下对这些估计量的性能进行了评估：(1)简单随机抽样(SRS)和(2)两阶段分层调查。随着倾向得分或结果模型误指定的程度增加，偏差和均方根误差也增加，而覆盖率则减少。简单随机样本和复杂调查设计的结果相似。
Model misspecification is a potential problem for any parametric-model based analysis. However, the measurement and consequences of model misspecification have not been well formalized in the context of causal inference. A measure of model misspecification is proposed, and the consequences of model misspecification in non-experimental causal inference methods are investigated. The metric is then used to explore which estimators are more sensitive to misspecification of the outcome and/or treatment assignment model. Three frequently used estimators of the treatment effect are considered, all of which rely on the propensity score: (1) full matching, (2) 1:1 nearest neighbor matching, and (3) weighting. The performance of these estimators is evaluated under two different sampling designs: (1) simple random sampling (SRS) and (2) a two-stage stratified survey. As the degree of misspecification of either the propensity score or outcome model increases, so does the bias and the root mean square error, while the coverage decreases. Results are similar for the simple random sample and a complex survey design.
登录
查看更多内容
DOI：
10.1198/016214504000001187
发表时间：
2004-09-01
影响因子：
3.7
作者：
Imai, K;van Dyk, DA
通讯作者：
van Dyk, DA
影响因子：
6.1
作者：
Hahn, JY
通讯作者：
Hahn, JY
影响因子：
3.7
作者：
ROSENBAUM, PR;RUBIN, DB
通讯作者：
RUBIN, DB
影响因子：
1.4
作者：
Ridgeway G;Kovalchik SA;Griffin BA;Kabeto MU
通讯作者：
Kabeto MU
影响因子：
2.7
作者：
CARPENTER, RG
通讯作者：
CARPENTER, RG