Additive, multiplicative, and other models for disease risks.
Additive, multiplicative, and other models for disease risks.
复制标题
疾病风险的加法、乘法和其他模型。
DOI：
10.1093/oxfordjournals.aje.a112629
复制
发表时间：
1978
影响因子：
5
通讯作者：
T. Holford
中科院分区：
文献类型：
--
作者：
S. D. Walter;T. Holford
在最近的两篇论文中，Rothman（1,2）讨论了流行病学数据的因果分析方面。第一篇论文的主旨是，疾病风险的加性模型是唯一允许对所涉及的各种风险因素的因果性质进行充分评估的模型；相比之下，其他模型被描述为使用“任意”转换，这些转换“可能错误地暗示或掩盖相互作用的存在”，罗斯曼将相互作用用作因果关系的主要指标。一个具体的例子是，该模型将独立性视为相对风险的倍增，它不会“为相关因果机制提供线索”。第二篇论文(2)通过推导基于加性模型的相互作用指数来发展这一主题。在我们看来，Rothman的立场是站不住脚的，不能作为流行病学数据分析的独家策略。我们将论证，在缺乏特定生理模型的情况下，相互作用的定义确实是任意的，定义的选择应该在某种程度上取决于
In two recent papers, Rothman (1, 2) discusses aspects of causal analysis of epidemiologic data. The main thrust of the first paper is that an additive model for disease risks is the only one which allows an adequate evaluation of the causal nature of various risk factors involved; in contrast, other models are described as using" arbitrary" transformations which" can falsely suggest or mask the presence of interaction", interaction being used by Rothman as the primary indicator of causality. A specific example, the model which regards independence as a multiplication of relative risks, will not" provide clues to the causal mechanisms involved". The second paper (2) develops this theme by deriving an index of interaction based on the additive model.In our view, Rothman's position is untenable as an exclusive strategy for the analysis of epidemiologic data. We will argue that the definition of interaction, in the absence of a specific physiologic model, is indeed arbitrary, and that the choice of the definition should depend to some extent on