当今健康医疗大数据及精准个性化医学时代，观察性研究备受青睐。然而，因其缺乏随机化，难以避免混杂偏倚(尤其是大量未测量或不可观察混杂因子所致的偏倚)，从而歪曲因果效应甚至导致错误结论。理论上倾向性得分、工具变量等可控制混杂，但实践中仍面临种种挑战。本项目在结构化的因果图模型指导下，巧妙利用大多数观察性研究所具备的纵向“时间差”或横断面“空间差”信息，在时间维度上设置“后置辅助变量”或在空间维度设置“外置辅助变量”；进而，仅用暴露与结局的测量数据自身，在反事实因果框架下运用常规回归或相关策略就能准确估计并检验暴露对结局的因果效应。预期目标：构建观察性研究中混杂控制的后置/外置辅助变量因果推断理论方法体系。实现目标，不仅能有效控制纵向观察性研究、功能磁共振等实时检测、肿瘤等横断面组学检测数据分析中的混杂偏倚而逼近因果效应；而且，将丰富发展因果推断理论方法，发挥观察性研究在生物医学研究中的优势。

This is the era of health care big data and precision medicine, observational studies are specially favored. However, because of the lack of randomness, it is difficult to avoid the confounding bias (especially the bias caused by a large number of unmeasured or unobservable confounding factors), which distort the causal effect and even lead to erroneous conclusions. Theoretically, methods such as propensity score and instrumental variables can be used to control bias, but challenges still exist in practice. Under the guidance of structured causal graphic model theory, this project utilizes the information of "temporal difference" in longitudinal study or "spatial difference" in cross-sectional study, and set "Postpositive auxiliary variable" in temporal dimension or "Extrapositive auxiliary variable" in spatial dimension; then, the causal effect of exposure to outcome can be accurately estimated by using traditional regression or related strategies under the framework of counterfactual causality, with only exposure and outcome data themselves. Anticipated Objective: To construct a theoretical system of causal inference of postpositive/extrapositive auxiliary variables for controlling confounding bias in observational studies. The realization of the goal can not only effectively control the confounding bias in longitudinal observation, real-time detection of functional magnetic resonance, cross-sectional omics data analysis of tumor, but also approach causal effect; in addition, it will enrich the causal inference theory and make the observational studies play an important role in biomedical research.