Causal Inference in Observational Studies

观察研究中的因果推理

• 批准号: 1260782
• 负责人: Dylan Small
• 金额: $ 29.66万
• 依托单位: University of Pennsylvania
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2013
• 资助国家: 美国
• 起止时间: 2013-03-15 至 2016-02-29
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1260782&HistoricalAwards=false
• 关键词: Causal; Inference; Observational; Studies

An observational study is an empiric investigation of the effects of a treatment, policy, intervention, or exposure that was not randomly assigned to subjects as it would be in a randomized experiment. Observational studies are common in most fields that study people because harmful or unwanted treatments cannot be imposed on human subjects for experimental purposes. The central difficulty in an observational study is that, because treatments were not randomly assigned, the subjects receiving different treatments may not be comparable, so differing outcomes after treatment may not be effects caused by the treatment. If the treatment groups differ before treatment in ways that have been measured, there is an overt bias that often can be removed by adjustments, such as matching. There often is concern that treatment groups differed before treatment in ways that were not measured, that is, concern about hidden biases. This research project will focus on hidden biases that generally cannot be removed by adjustments and must be addressed by other means. Prior work has shown that the design of an observational study strongly affects its sensitivity to hidden biases. This project is comprised of four components. The first concerns the possibility that effect modification (a treatment by covariate interaction) may under suitable circumstances be exploited to reduce sensitivity to unmeasured biases. The second component concerns the relationship between case definition in case-control studies and sensitivity to unmeasured biases. The third concerns clustered treatment assignment and sensitivity to unmeasured biases. The final component concerns use of risk set matching to provide finer control of time dependent instrumental variables.Observational studies are common in economics, education, epidemiology, medicine, public policy, and sociology. Improved methodology for observational studies has the potential to lead to improved policies and practices of both public and private institutions.

观察性研究是对治疗、政策、干预或暴露的影响进行的经验性研究，这些研究不是像随机实验那样随机分配给受试者。 观察性研究在大多数研究人类的领域都很常见，因为不能为了实验目的而对人类受试者施加有害或不必要的治疗。 观察性研究的主要困难在于，由于治疗不是随机分配的，接受不同治疗的受试者可能不具有可比性，因此治疗后的不同结局可能不是治疗引起的效应。 如果治疗组在治疗前的测量方式不同，则存在明显的偏差，通常可以通过调整（例如匹配）来消除。 人们常常担心，治疗组在治疗前的差异是无法测量的，也就是说，担心隐藏的偏见。 该研究项目将侧重于隐藏的偏见，这些偏见通常无法通过调整来消除，必须通过其他手段加以解决。 先前的研究表明，观察性研究的设计强烈影响其对隐藏偏倚的敏感性。 该项目由四个部分组成。 第一个问题是，在适当的情况下，可以利用效应修正（协变量相互作用的治疗）来降低对不可测偏倚的敏感性。 第二个组成部分涉及病例对照研究中的病例定义与对不可测量偏差的敏感性之间的关系。 第三个问题涉及聚类治疗分配和对未测量偏倚的敏感性。 最后一个部分涉及使用风险集匹配来提供对时间依赖工具变量的更精细的控制。观察性研究在经济学、教育、流行病学、医学、公共政策和社会学中很常见。 改进观察性研究的方法有可能改进公共和私营机构的政策和做法。