Shahar Eyal
Division of Epidemiology and Biostatistics, Mel and Enid Zuckerman College of Public Health, University of Arizona, Tucson, AZ 85724, USA.
J Eval Clin Pract. 2007 Oct;13(5):814-6. doi: 10.1111/j.1365-2753.2007.00796.x.
In recent years, numerous research methodologists have argued forcefully that any estimated effect from an observational study or a randomized trial should apply to a 'target population'- to a finite group of people. Some methods to adjust for confounders heavily draw upon this idea.
I cite a recently published paper in The American Journal of Epidemiology that linked methods to adjust for confounders to the concept of a 'target population'. I explain that the requirement to specify a finite population as the target of causal inference is rooted in two extreme models of causation: determinism and stochastic causation.
I argue that the 'target population' epistemology is scientifically irrelevant and so are methods to handle confounders that are founded on this paradigm, namely, standardization, inverse-probability-of-treatment weighting and SMR-weighting. Finally, I propose a simple alternative framework under an indeterministic model of causation. According to my proposed model, a causal parameter is not tied to any finite population and its estimate is a (fallible) scientific conjecture about a homogeneous, individual-level effect.
近年来，众多研究方法学家有力地指出，观察性研究或随机试验得出的任何估计效应都应适用于“目标人群”——一个有限的人群组。一些调整混杂因素的方法很大程度上借鉴了这一理念。
我引用了最近发表在《美国流行病学杂志》上的一篇论文，该论文将调整混杂因素的方法与“目标人群”的概念联系起来。我解释说，将有限人群指定为因果推断目标的要求源于两种极端的因果模型：决定论和随机因果关系。
我认为“目标人群”认识论在科学上是无关紧要的，基于这一范式处理混杂因素的方法也是如此，即标准化、治疗逆概率加权和标准化死亡比加权。最后，我在非确定性因果模型下提出了一个简单的替代框架。根据我提出的模型，因果参数不与任何有限人群相关联，其估计是关于同质个体水平效应的（可能有误的）科学推测。
