Causal Inference with Massive and Complex data: High-dimensionality and Network Interference

现代观测数据库对新发现大有希望。迄今为止，通过在大量和复杂的数据集中发现关联已经取得了许多进展。寻找关联模式是重要的第一步，但它本身并不能带来可操作的见解。这促使我们提出的方案超越关联，在“大数据”环境中评估因果关系，以便我们更好地理解大自然的无限复杂性，并利用这些知识做出更好的决策。具体而言，我们专注于解决当代大规模和复杂数据集带来的两个关键挑战：高维和复杂的网络结构。我们将开发一套新的方法，由数据密集型领域的应用驱动，如神经成像、遗传学和网络分析。这些进步将推动因果推理超越传统设置的前沿，同时将因果思维带入大数据分析。该项目还将为有前途的本科生和研究生以及博士后研究人员提供培训机会，特别是那些来自历史上代表性不足的群体的研究人员。通过这项资助开发的方法将用R编码，并与相关研究报告一起公开发布。

Modern observational databases hold great promise for new discoveries. To date, a lot of advances have been made by finding associations in massive and complex data sets. Finding patterns of associations is an important first step, but does not by itself lead to actionable insights. This motivates the proposed program to go beyond association and assess causation in "big data" settings, so that we can better understand the infinite complexity of nature, and use such knowledge to make better decisions. Specifically we focus on addressing two key challenges posed by contemporary massive and complex data sets: high-dimensionality and complex network structure. We are going to develop a novel set of methods driven by applications in data-intensive fields such as neuroimaging, genetics and network analysis. These advances will push the frontiers of causal inference beyond conventional settings, and at the same time bring causal thinking to big data analytics. This program will also provide training opportunities for promising undergraduate and graduate students as well as postdoctoral researchers, especially those from historically underrepresented groups. Methods developed through this grant will be coded in R and made publicly available along with publication of related research reports.